Taken by tokens OpenAIs new AI image generator is potent and bound to provoke The visual apocalypse is probably nigh, but perhaps seeing was never believing. Benj Edwards Mar 27, 2025 7:15 am | 42 A trio of AI-generated images created using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI A trio of AI-generated images created using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreThe arrival of OpenAI's DALL-E 2 in the spring of 2022 marked a turning point in AI when text-to-image generation suddenly became accessible to a select group of users, creating a community of digital explorers who experienced wonder and controversy as the technology automated the act of visual creation.But like many early AI systems, DALL-E 2 struggled with consistent text rendering, often producing garbled words and phrases within images. It also had limitations in following complex prompts with multiple elements, sometimes missing key details or misinterpreting instructions. These shortcomings left room for improvement that OpenAI would address in subsequent iterations, such as DALL-E 3 in 2023.On Tuesday, OpenAI announced new multimodal image generation capabilities that are directly integrated into its GPT-4o AI language model, making it the default image generator within the ChatGPT interface. The integration, called "4o Image Generation" (which we'll call "4o IG" for short), allows the model to follow prompts more accurately (with better text rendering than DALL-E 3) and respond to chat context for image modification instructions. An AI-generated cat in a car drinking a can of beer created by OpenAI's 4o Image Generation model. OpenAI An AI-generated cat in a car drinking a can of beer created by OpenAI's 4o Image Generation model. OpenAI An AI-generated photo of Abraham Lincoln holding an Ars Technica sign created by OpenAI's 4o Image Generation model. OpenAI An AI-generated photo of Abraham Lincoln holding an Ars Technica sign created by OpenAI's 4o Image Generation model. OpenAI An AI-generated image of "a muscular barbarian with weapons beside a CRT television set, cinematic, 8K, studio lighting" created by OpenAI's 4o Image Generation model. An AI-generated image of "a muscular barbarian with weapons beside a CRT television set, cinematic, 8K, studio lighting" created by OpenAI's 4o Image Generation model.An AI-generated photo of Abraham Lincoln holding an Ars Technica sign created by OpenAI's 4o Image Generation model. OpenAI An AI-generated image of "a muscular barbarian with weapons beside a CRT television set, cinematic, 8K, studio lighting" created by OpenAI's 4o Image Generation model. An AI-generated "Queen of the Universe" by OpenAI's 4o Image Generation model. OpenAI An AI-generated plate of pickles created by OpenAI's 4o Image Generation model. OpenAI Generating a gaming PC with 1000 RGB lights using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Generating a flaming cheeseburger using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards The new image generation feature began rolling out Tuesday to ChatGPT Free, Plus, Pro, and Team users, with Enterprise and Education access coming later. The capability is also available within OpenAI's Sora video generation tool. OpenAI told Ars that the image generation when GPT-4.5 is selected calls upon the same 4o-based image generation model as when GPT-4o is selected in the ChatGPT interface.Like DALL-E 2 before it, 4o IG is bound to provoke debate as it enables sophisticated media manipulation capabilities that were once the domain of sci-fi and skilled human creators into an accessible AI tool that people can use through simple text prompts. It will also likely ignite a new round of controversy over artistic styles and copyrightbut more on that below. 4o IG can change our perception of media reality. Given this actual photo of a dog... Benj Edwards 4o IG can change our perception of media reality. Given this actual photo of a dog... Benj Edwards ...the AI model can change what the dog is doing in a realistic way, such as playing with a fictional puppy inserted into the scene. OpenAI / Benj Edwards ...the AI model can change what the dog is doing in a realistic way, such as playing with a fictional puppy inserted into the scene. OpenAI / Benj Edwards 4o IG can change our perception of media reality. Given this actual photo of a dog... Benj Edwards ...the AI model can change what the dog is doing in a realistic way, such as playing with a fictional puppy inserted into the scene. OpenAI / Benj Edwards Some users on social media initially reported confusion since there's no UI indication of which image generator is active, but you'll know it's the new model if the generation is ultra slow and proceeds from top to bottom. The previous DALL-E model remains available through a dedicated "DALL-E GPT" interface, while API access to GPT-4o image generation is expected within weeks.Truly multimodal output4o IG represents a shift to "native multimodal image generation," where the large language model processes and outputs image data directly as tokens. That's a big deal, because it means image tokens and text tokens share the same neural network. It leads to new flexibility in image creation and modification.Despite baking-in multimodal image generation capabilities when GPT-4o launched in May 2024when the "o" in GPT-4o was touted as standing for "omni" to highlight its ability to both understand and generate text, images, and audioOpenAI has taken over 10 months to deliver the functionality to users, despite OpenAI president Greg Brock teasing the feature on X last year.OpenAI was likely goaded by the release of Google's multimodal LLM-based image generator called "Gemini 2.0 Flash (Image Generation) Experimental," last week. The tech giants continue their AI arms race, with each attempting to one-up the other.And perhaps we know why OpenAI waited: At a reasonable resolution and level of detail, the new 4o IG process is extremely slow, taking anywhere from 30 seconds to one minute (or longer) for each image. Generating a four-panel comic using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Generating a four-panel comic using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Giving the man in the four-panel comic a beard using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Giving the man in the four-panel comic a beard using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Generating a four-panel comic using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Giving the man in the four-panel comic a beard using OpenAI's 4o Image Generation model in ChatGPT. OpenAI / Benj Edwards Even if it's slow (for now), the ability to generate images using a purely autoregressive approach is arguably a major leap for OpenAI due to its flexibility. But it's also very compute-intensive, since the model generates the image token by token, building it sequentially. This contrasts with diffusion-based methods like DALL-E 3, which start with random noise and gradually refine an entire image over many iterative steps.Conversational image editingIn a blog post, OpenAI positions 4o Image Generation as moving beyond generating "surreal, breathtaking scenes" seen with earlier AI image generators and toward creating "workhorse imagery" like logos and diagrams used for communication.The company particularly notes improved text rendering within images, a capability where previous text-to-image models often spectacularly failed, often turning "Happy Birthday" into something resembling alien hieroglyphics.OpenAI claims several key improvements: users can refine images through conversation while maintaining visual consistency; the system can analyze uploaded images and incorporate their details into new generations; and it offers stronger photorealismalthough what constitutes photorealism (for example, imitations of HDR camera features, detail level, and image contrast) can be subjective. A screenshot of OpenAI's 4o Image Generation model in ChatGPT. We see an existing AI-generated image of a barbarian and a TV set, then a request to set the TV set on fire. Credit: OpenAI / Benj Edwards In its blog post, OpenAI provided examples of intended uses for the image generator, including creating diagrams, infographics, social media graphics using specific color codes, logos, instruction posters, business cards, custom stock photos with transparent backgrounds, editing user photos, or visualizing concepts discussed earlier in a chat conversation.Notably absent: Any mention of the artists and graphic designers whose jobs might be affected by this technology. As we covered throughout 2022 and 2023, job impact is still a top concern among critics of AI-generated graphics.Fluid media manipulationShortly after OpenAI launched 4o Image Generation, the AI community on X put the feature through its paces, finding that it is quite capable at inserting someone's face into an existing image, creating fake screenshots, and converting meme photos into the style of Studio Ghibli, South Park, felt, Muppets, Rick and Morty, Family Guy, and much more.It seems like we're entering a completely fluid media "reality" courtesy of a tool that can effortlessly convert visual media between styles. The styles also potentially encroach upon protected intellectual property. Given what Studio Ghibli co-founder Hayao Miyazaki has previously said about AI-generated artwork ("I strongly feel that this is an insult to life itself."), it seems he'd be unlikely to appreciate the current AI-generated Ghibli fad on X at the moment. An Internet meme converted into "Studio Ghibli" style art by people on X. OpenAI / BarseeAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Barsee An Internet meme converted into "Studio Ghibli" style art by people on X. OpenAI / Justine MooreAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Justine MooreAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / BarseeAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Justine Moore An Internet meme converted into "Studio Ghibli" style art by people on X. OpenAI / Justine MooreAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Justine Moore An Internet meme converted into "Studio Ghibli" style art by people on X. OpenAI / Manuel CalaveraAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Manuel CalaveraAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Justine MooreAn Internet meme converted into "Studio Ghibli" style art by people on X.OpenAI / Manuel CalaveraTo get a sense of what 4o IG can do ourselves, we ran some informal tests, including some of the usual CRT barbarians, queens of the universe, and beer-drinking cats, which you've already seen above (and of course, the plate of pickles.)The ChatGPT interface with the new 4o image model is conversational (like before with DALL-E 3), but you can suggest changes over time. For example, we took the author's EGA pixel bio (as we did with Google's model last week) and attempted to give it a full body. Arguably, Google's more limited image model did a far better job than 4o IG. Giving the author's pixel avatar a body using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI / Benj Edwards While my pixel avatar was commissioned from the very human (and talented) Julia Minamata in 2020, I also tried to convert the inspiration image for my avatar (which features me and legendary video game engineer Ed Smith) into EGA pixel style to see what would happen. In my opinion, the result proves the continued superiority of human artistry and attention to detail. Converting a photo of Benj Edwards and video game legend Ed Smith into "EGA pixel art" using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI / Benj Edwards We also tried to see how many objects 4o Image Generation could cram into an image, inspired by a 2023 tweet by Nathan Shipley when he was evaluating DALL-E 3 shortly after its release. We did not account for every object, but it looks like most of them are there. Generating an image of a surfer holding tons of items, inspired by a 2023 Twitter post from Nathan Shipley. Credit: OpenAI / Benj Edwards On social media, other people have manipulated images using 4o IG (like Simon Willison's bear selfie), so we tried changing an AI-generated note featured in an article last year. It worked fairly well, though it did not really imitate the handwriting style as requested. Modifying text in an image using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI / Benj Edwards To take text generation a little further, we generated a poem about barbarians using ChatGPT, then fed it into an image prompt. The result feels roughly equivalent to diffusion-based Flux in capabilitymaybe slightly betterbut there are still some obvious mistakes here and there, such as repeated letters. Testing text generation using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI / Benj Edwards We also tested the model's ability to create logos featuring our favorite fictional Moonshark brand. One of the logos not pictured here was delivered as a transparent PNG file with an alpha channel. This may be a useful capability for some people in a pinch, but to the extent that the model may produce "good enough" (not exceptional, but looks OK at a glance) logos for the price of $o (not including an OpenAI subscription), it may end up competing with some human logo designers, and that will likely cause some consternation among professional artists. Generating a "Moonshark Moon Pies" logo using OpenAI's 4o Image Generation model in ChatGPT. Credit: OpenAI / Benj Edwards Frankly, this model is so slow we didn't have time to test everything before we needed to get this article out the door. It can do much more than we have shown heresuch as adding items to scenes or removing them. We may explore more capabilities in a future article.LimitationsBy now, you've seen that, like previous AI image generators, 4o IG is not perfect in quality: It consistently renders the author's nose at an incorrect size.Other than that, while this is one of the most capable AI image generators ever created, OpenAI openly acknowledges significant limitations of the model. For example, 4o IG sometimes crops images too tightly or includes inaccurate information (confabulations) with vague prompts or when rendering topics it hasn't encountered in its training data.The model also tends to fail when rendering more than 1020 objects or concepts simultaneously (making tasks like generating an accurate periodic table currently impossible) and struggles with non-Latin text fonts. Image editing is currently unreliable over many multiple passes, with a specific bug affecting face editing consistency that OpenAI says it plans to fix soon. And it's not great with dense charts or accurately rendering graphs or technical diagrams. In our testing, 4o Image Generation produced mostly accurate but flawed electronic circuit schematics.Move fast and break everythingEven with those limitations, multimodal image generators are an early step into a much larger world of completely plastic media reality where any pixel can be manipulated on demand with no particular photo editing skill required. That brings with it potential benefits, ethical pitfalls, and the potential for terrible abuse.In a notable shift from DALL-E, OpenAI now allows 4o IG to generate adult public figures (not children) with certain safeguards, while letting public figures opt out if desired. Like DALL-E, the model still blocks policy-violating content requests (such as graphic violence, nudity, and sex).The ability for 4o Image Generation to imitate celebrity likenesses, brand logos, and Studio Ghibli films reinforces and reminds us how GPT-4o is partly (aside from some licensed content) a product of a massive scrape of the Internet without regard to copyright or consent from artists. That mass-scraping practice has resulted in lawsuits against OpenAI in the past, and we would not be surprised to see more lawsuits or at least public complaints from celebrities (or their estates) about their likenesses potentially being misused.On X, OpenAI CEO Sam Altman wrote about the company's somewhat devil-may-care position about 4o IG: "This represents a new high-water mark for us in allowing creative freedom. People are going to create some really amazing stuff and some stuff that may offend people; what we'd like to aim for is that the tool doesn't create offensive stuff unless you want it to, in which case within reason it does." An original photo of the author beside AI-generated images created by OpenAI's 4o Image Generation model. From second left to right: Studio Ghibli style, Muppet style, and pasta style. Credit: OpenAI / Benj Edwards Zooming out, GPT-4o's image generation model (and the technology behind it, once open source) feels like it further erodes trust in remotely produced media. While we've always needed to verify important media through context and trusted sources, these new tools may further expand the "deep doubt" media skepticism that's become necessary in the age of AI. By opening up photorealistic image manipulation to the masses, more people than ever can create or alter visual media without specialized skills.While OpenAI includes C2PA metadata in all generated images, that data can be stripped away and might not matter much in the context of a deceptive social media post. But 4o IG doesn't change what has always been true: We judge information primarily by the reputation of its messenger, not by the pixels themselves. Forgery existed long before AI. It reinforces that everyone needs media literacy skillsunderstanding that context and source verification have always been the best arbiters of media authenticity.For now, Altman is ready to take on the risks of releasing the technology into the world. "As we talk about in our model spec, we think putting this intellectual freedom and control in the hands of users is the right thing to do, but we will observe how it goes and listen to society," Altman wrote on X. "We think respecting the very wide bounds society will eventually choose to set for AI is the right thing to do, and increasingly important as we get closer to AGI. Thanks in advance for the understanding as we work through this."Benj EdwardsSenior AI ReporterBenj EdwardsSenior AI Reporter Benj Edwards is Ars Technica's Senior AI Reporter and founder of the site's dedicated AI beat in 2022. He's also a tech historian with almost two decades of experience. In his free time, he writes and records music, collects vintage computers, and enjoys nature. He lives in Raleigh, NC. 42 Comments

Traffic safety After 50 million miles, Waymos crash a lot less than human drivers Waymo has been in dozens of crashes. Most were not Waymo's fault. Timothy B. Lee Mar 27, 2025 7:00 am | 26 A driverless Waymo in Los Angeles. Credit: P_Wei via Getty A driverless Waymo in Los Angeles. Credit: P_Wei via Getty Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreThe first ever fatal crash involving a fully driverless vehicle occurred in San Francisco on January 19. The driverless vehicle belonged to Waymo, but the crash was not Waymos fault.Heres what happened: A Waymo with no driver or passengers stopped for a red light. Another car stopped behind the Waymo. Then, according to Waymo, a human-driven SUV rear-ended the other vehiclesat high speed, causing a six-car pileup that killed one person and injured five others. Someones dog also died in the crash.Another major Waymo crash occurred in October in San Francisco. Once again, a driverless Waymo was stopped for a red light. According to Waymo, a vehicle traveling in the opposite direction crossed the double yellow line and crashed into an SUV that was stopped to the Waymos left. The force of the impact shoved the SUV into the Waymo. One person was seriously injured.These two incidents produced worse injuries than any other Waymo crash in the last nine months. But in other respects, they were typical Waymo crashes. Most Waymo crashes involve a Waymo vehicle scrupulously following the rules while a human driver flouts them, speeding, running red lights, careening out of their lanes, and so forth.Waymos service will only grow in the coming months and years. So Waymo will inevitably be involved in more crashesincluding some crashes that cause serious injuries and even death.But as this happens, its crucial to keep the denominator in mind. Since 2020, Waymo has reported roughly 60 crashes serious enough to trigger an airbag or cause an injury. But those crashes occurred over more than 50 million miles of driverless operations. If you randomly selected 50 million miles of human drivingthats roughly 70 lifetimes behind the wheelyou would likely see far more serious crashes than Waymo has experienced to date.Federal regulations require Waymo to report all significant crashes, whether or not the Waymo vehicle was at faultindeed, whether or not the Waymo is even moving at the time of the crash. Ive spent the last few days poring over Waymos crash reports from the last nine months. Lets dig in.Examining Waymos safety record since JulyLast September, IanalyzedWaymo crashes through June 2024. So this section will focus on crashes between July 2024 and February 2025. During that period, Waymo reported 38 crashes that were serious enough to either cause an (alleged) injury or an airbag deployment.In my view, only one of these crashes was clearly Waymos fault. Waymo may have been responsible for three other crashesthere wasnt enough information to say for certain. The remaining 34 crashes seemed to be mostly or entirely the fault of others:The two serious crashes I mentioned at the start of this article are among 16 crashes where another vehicle crashed into a stationary Waymo (or caused a multi-car pileup involving a stationary Waymo). This included 10 rear-end crashes, three side-swipe crashes, and three crashes where a vehicle coming from the opposite direction crossed the center line.Anothereight crashesinvolved another car (or in one case a bicycle) rear-ending a moving Waymo.A furtherfive crashesinvolved another vehicle veering into a Waymos right of way. This included a car running a red light, a scooter running a red light, and a car running a stop sign.Three crashesoccurred while Waymo was dropping a passenger off. The passenger opened the door and hit a passing car or bicycle. Waymo has aSafe Exit programto alert passengers and prevent this kind of crash, but its not foolproof.There were two incidents where it seems like no crash happened at all:In one incident, Waymo says that its vehicle slowed and moved slightly to the left within its lane, preparing to change lanes due to a stopped truck ahead. This apparently spooked an SUV driver in the next lane, who jerked the wheel to the left and ran into the opposite curb. Waymo says its vehicle never left its lane or made contact with the SUV.In another incident, a pedestrian walked in front of a stopped Waymo. The Waymo began moving after the pedestrian had passed, but then the pedestrian turned around and approached the Waymo AV. According to Waymo, the pedestrian may have made contact with the driver side of the Waymo AV and later claimed to have a minor injury. Waymos report stops just short of calling this pedestrian a liar.So thats a total of 34 crashes. I dont want to make categorical statements about these crashes because in most cases, I only have Waymos side of the story. But it doesnt seem like Waymo was at fault in any of them.There was one crash where Waymo clearly seemed to be at fault: In December, a Waymo in Los Angeles ran into a plastic crate, pushing it into the path of a scooter in the next lane. The scooterist hit the crate and fell down. Waymo doesnt know whether the person riding the scooter was injured.I had trouble judging the final three crashes, all of which involved another vehicle making an unprotected left turn across a Waymos lane of travel. In two of these cases, Waymo says its vehicle slammed on the brakes but couldnt stop in time to avoid a crash. In the third case, the other vehicle hit the Waymo from the side. Waymos summaries make it sound like the other car was at fault in all three cases, but I dont feel like I have enough information to make a definite judgment.Even if we assume all three of these crashes were Waymos fault, that would still mean that a large majority of the 38 serious crashes were not Waymos fault. And as well see, Waymo vehicles are involved in many fewer serious crashes than human-driven vehicles.Waymos get in fewer crashes than human driversAnother way to evaluate the safety of Waymo vehicles is by comparing their per-mile crash rate to human drivers. Waymo has been regularly publishing data about this over the last couple of years. Its most recent release came last week, when Waymo updated itssafety data hubto cover crashes through the end of 2024.Waymo knows exactly how many times its vehicles have crashed. Whats tricky is figuring out the appropriate human baseline, since human drivers dont necessarily report every crash. Waymo has tried to address this by estimating human crash rates in its two biggest marketsPhoenix and San Francisco. Waymos analysis focused on the 44 million miles Waymo had driven in these cities through December, ignoring its smaller operations in Los Angeles and Austin.Using human crash data, Waymo estimated that human drivers on the same roads would get into 78 crashes serious enough to trigger an airbag. By comparison, Waymos driverless vehicles only got into 13 airbag crashes. That represents an83 percent reductionin airbag crashes relative to typical human drivers.This is slightly worse than last September, when Waymo estimated an 84 percent reduction in airbag crashes over Waymos first 21 million miles.Over the same 44 million miles, Waymo estimates that human drivers would get into 190 crashes serious enough to cause an injury. Instead, Waymo only got in 36 injury-causing crashes across San Francisco or Phoenix. Thats an81 percent reductionin injury-causing crashes.This is a significant improvement over last September, when Waymo estimated its cars had 73 percent fewer injury-causing crashes over its first 21 million driverless miles.Insurance claims against Waymo are about 90 percent lowerThe above analysis counts all crashes, whether or not Waymos technology was at fault. Things look even better for Waymo if we focus on crashes where Waymo was determined to be responsible for a crash.To assess this, Waymoco-authored a studyin December with the insurance giant Swiss Re. It focused on crashes that led to successful insurance claims against Waymo. This data seems particularly credible because third parties, not Waymo, decide when a crash is serious enough to file an insurance claim. And claims adjusters, not Waymo, decide whether to hold Waymo responsible for a crash.But one downside is that it takes a few months for insurance claims to be filed. So the December report focused on crashes that occurred through July 2024.Waymo had completed 25 million driverless miles by July 2024. And by the end of November 2024, Waymo had faced only two potentially successful claims for bodily injury. Both claims are pending, which means they could still be resolved in Waymos favor.One of them was this crash that I described at the beginning of mySeptember articleabout Waymos safety record:On a Friday evening last November, police chased a silver sedan across the San Francisco Bay Bridge. The fleeing vehicle entered San Francisco and went careening through the citys crowded streets. At the intersection of 11th and Folsom streets, it sideswiped the fronts of two other vehicles, veered onto a sidewalk, and hit two pedestrians.According to a local news story, both pedestrians were taken to the hospital, with one suffering major injuries. The driver of the silver sedan was injured, as was a passenger in one of the other vehicles. No one was injured in the third car, a driverless Waymo robotaxi.It seems unlikely that an insurance adjuster will ultimately hold Waymo responsible for these injuries.The other pending injury claim doesnt seem like a slam dunk, either. In that case, another vehicle steered into a bike lane before crashing into a Waymo as it was making a left turn.But lets assume that both crashes are judged to be Waymos fault. That would still be a strong overall safety record.Based on insurance industry records, Waymo and Swiss Re estimate that human drivers in San Francisco and Phoenix would generate about 26 successful bodily injury claims over 25 million miles of driving. So even if both of the pending claims against Waymo succeed, two injuries represent a more than 90 percent reduction in successful injury claims relative to typical human drivers.The reduction in property damage claims is almost as dramatic. Waymos vehicles generated nine successful or pending property damage claims over its first 25 million miles. Waymo and Swiss Re estimate that human drivers in the same geographic areas would have generated 78 property damage claims. So Waymo generated 88 percent fewer property damage claims than typical human drivers.Timothy B. Lee was on staff at Ars Technica from 2017 to 2021. Today he writes Understanding AI, a newsletter that explores how AI works and how it's changing our world. You can subscribe here.Timothy B. LeeSenior tech policy reporterTimothy B. LeeSenior tech policy reporter Timothy is a senior reporter covering tech policy and the future of transportation. He lives in Washington DC. 26 Comments

Lisa Morgan, Freelance WriterMarch 27, 20258 Min ReadNils Ackermann via Alamy StockData silos have been plaguing organizations since before the data analytics gold rush. Sadly, data silos remain an issue in many organizations, which calls into question the reliability of AI outputs.Data silos are making it much harder for agents to get unified insights based on a holistic view of the data about an object of interest, such as a customer or an employee, or just a single user, says Michael Berthold, CEO and co-founder of data analytics platform provider KNIME. For example, agents struggle with isolated data sources, [like] a human having to go to the CRM to see information about a company and the current contract history, then go to the support system to find out more about ongoing technical issues, and then also check the online forum to see if employees of the customer posted something there.According to a recent Gartner survey, 63% of organizations either do not have or are unsure if they have the rightdatamanagement practices for AI. In fact, Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported byAI-ready data.How Data Silos Form and What to Do About ThemTool vendors are trying to make the flow of data between systems easier by providing integrations with other tools. Similarly, an agent will benefit from having one place to go to get information about a customer.Related:Michael Berthold, KNIMEMichael Berthold, KNIMEIn an ideal world, all data would be integrated. That was the promise of data warehouses years ago, and its still what is being promised. Especially companies with more legacy data and systems will continue to have data silos, says Berthold.AI models require high-quality data to deliver optimal performance. Poor data leads to underperforming models, which can cost organizations tens of millions of dollars or more, according to Gordon Robinson, senior director, data management R&D at data and AI solution provider SAS.Inconsistent data across silos means different parts of an organization may track similar data independently, leading to discrepancies and the lack of a single source of truth, says Robinson. Data silos also can lead to incomplete AI model training. When AI models are trained on fragmented data rather than a comprehensive dataset, they fail to reach their full potential and deliver optimal insights.Josh Weinick, a sales engineer at AI-powered cybersecurity automation platform Blink Ops has seen cases in which a chatbot is unable to provide accurate customer support because it doesnt have access to sales or product data living in another departments separate system.Related:Most silos are caused by a mix of legacy infrastructure, organizational culture and inconsistent data standards. When teams cling to their own systems and definitions, or when older technology doesnt integrate well with modern AI platforms, its easy for silos to form, says Weinick. Mergers and acquisitions can also play a role. Newly acquired business units often bring their own tech stacks, which stay isolated unless leadership prioritizes integration.Without leadership buy-in and a culture of data sharing, departments tend to guard their data.Ashwin Rajeeva, co-founder and CTO at enterprise data observability company Acceldata says data silos restrict AIs access to complete, high-quality data, which leads to biased models, inconsistent insights and unreliable automation.Fragmented datasets make it difficult for AI agents to understand context, reducing their effectiveness in decision-making and business impact, says Rajeeva. Eliminating silos is essential for AI to scale, improve efficiency and deliver meaningful enterprise value.The root causes of the data access problem are legacy infrastructure, multi-cloud environments, decentralized data ownership and weak governance.Related:A data-first AI strategy focused on governance, interoperability, and observability is key. Enterprises should implement automated data quality checks, real-time monitoring and lineage tracking to ensure AI models operate on accurate, consistent data. Aligning data strategy with business objectives and fostering cross-functional collaboration accelerates AI adoption and impact, says Rajeeva.Gokul Naidu, senior manager at SAP says silos can cause gaps in model training and may require manual consolidation or cross-team requests.By the time information is merged, it may already become outdated, slowing the feedback loop for AI driven optimizations and reducing potential ROI, says Naidu. When I wear a FinOps hat I see that silos obscure the value of unit economics, such as cost per transaction, cost per user, and limit the ability to measure how each service or feature contributes to overall business value.In his view, cultural resistance to sharing, a lack of standards and governance, legacy apps and technical debt contribute to data fragmentation, making it difficult to establish a unified data strategy. To overcome them, he suggests doing the opposite, which is promoting a culture of sharing, having a unified data strategy, and using automation and observability.Paul Graeve, CEO at IT system data services provider The Data Group points to SaaS systems. Specifically, organizations are not investing the time, energy, and money necessary to load SaaS data into a data warehouse where the organization can own the data, clean it, and effectively use the data for any important business initiative.Your data is locked away in all these SaaS platforms scattered around the globe. This can be scary considering your data is your most valuable asset, says Graeve. The only way you can effectively and efficiently use your data for AI, analytics, portals -- for any initiative -- is to consolidate all your data into a one-version-of-the-truth data warehouse. Until you have your data in one place where you can see it, fix it, enrich it and efficiently use it, youre going to struggle successfully implementing any AI initiative.Paul Graeve, The Data GroupPaul Graeve, The Data GroupArmando Franco, director of business modernization services at TEKsystems Global Services, says data silos limit access to comprehensive training data, reducing model accuracy, and introducing inconsistencies due to conflicting governance and duplication. They also create inefficiencies in automation and decision-making, as AI agents require real-time access to unified data. Additionally, fragmented data poses security and compliance risks, potentially leading to regulatory violations if governance is not properly enforced.These challenges stem from outdated IT infrastructure, business unit fragmentation, and a lack of a unified data strategy, says Franco. Legacy systems were not designed for interoperability, while different departments using specialized tools create barriers to integration. Without centralized governance, enterprises struggle with inconsistent data management, and siloed AI initiatives lead to duplicated efforts and conflicting model outputs. Addressing these issues requires modernizing IT systems, fostering cross-team collaboration, and implementing a cohesive data strategy.Why Some Enterprises Struggle More Than OthersThe longer an organization exists, the more likely it is to be struggling with data silos.If a company has been around for a while, it will have different tools and systems, and the act of unifying it all is doomed from the start. Even worse, if that company bought a couple of other companies in recent years that brought along their own tools and data solutions, says KNIMEs Berthold. Dont dream of waiting for the famous data warehouse to solve everything. Dont try to put a bandage on the problem by starting to copy around data so it all creates a data swamp in one central location.Instead, its important to have a data integration, aggregation and analytics layer in place that allows everybody and AI agents to access a unified view. Berthold says organizations should ensure the technology in that layer is well-documented so future colleagues can understand its functionality and update it as data moves or new data sources are added.According to SAS Robinson, data silos within organizations often form around products or business functions, so many organizations still struggle to unlock the full potential of their data.The best way to overcome these challenges is by implementing a strong data governance framework within your organization. With increasing regulatory demands and the rising frequency and cost of data breaches, robust data governance is no longer a choice -- its a necessity, says Robinson. A successful data governance program starts with understanding what data you have, assessing its quality and tracking how it is used across the organization.Additionally, techniques like entity resolution can help create a single, unified view of data by integrating information from disparate silos into a centralized repository. However, many organizations have yet to invest in strong data governance. Meanwhile, AI governance is emerging as a crucial focus, especially as new AI regulations continue to evolve.Effective AI governance must be built on a solid foundation of robust data governance, says Robinson. If you havent invested in data governance or your current platform lacks robustness, this should be your top priority. Its no longer optional. Its a fundamental necessity for any data-driven organization today.In addition to that, Blink Ops Weinick says organizations should prepared to invest in modern data integration and metadata management and put strong security and governance frameworks in place from the start, so fears around compliance or breaches dont create massive delays.Most importantly, focus on cultivating a cross-functional mindset, says Weinick. Demonstrate quick wins by bringing together two siloed data sets to address a pressing business problem, then celebrate and scale those successes across the enterprise.About the AuthorLisa MorganFreelance WriterLisa Morgan is a freelance writer who covers business and IT strategy and emergingtechnology for InformationWeek. She has contributed articles, reports, and other types of content to many technology, business, and mainstream publications and sites including tech pubs, The Washington Post and The Economist Intelligence Unit. Frequent areas of coverage include AI, analytics, cloud, cybersecurity, mobility, software development, and emerging cultural issues affecting the C-suite.See more from Lisa MorganReportsMore ReportsNever Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.SIGN-UPYou May Also Like

Brian Bronson, Chief Executive Officer, Orion InnovationMarch 27, 20252 Min ReadAleksey Odintsov via Alamy StockWe need to do something with AI. This familiar refrain came from a client recently but with an unusual twist. Instead of developing their AI LLM as quickly as possible, they wanted to build a governance framework first to guide their AI initiative.This conversation sparked an important realization: Strategic AI governance, when done right, serves as a powerful technology enabler.The pressure to adopt AI has reached a fever pitch; organizations feel they dont exist if theyre not doing AI. However, this pressure often leads to rushed implementations that can damage a business or product roadmap rather than enhance it.A healthcare provider that we recently spoke with learned this the hard way.After rushing out with a new AI transcription system to keep pace with a competitor, the system had to be shut down when they discovered it inadvertently included sensitive patient information in meeting summaries.The takeaway: Successful AI implementations share a common thread. They treat governance as an acceleration framework, not an obstacle. This requires fundamentally rethinking how we approach technology governance.Strategic ReviewsTraditional yes/no governance approaches don't work for AI. A more effective strategy focuses on a development program that creates clear pathways to deployment based on risk levels and business impact. For example, projects using established AI models with limited customer impact can move through a rapid approval process, while those involving sensitive data or custom AI development receive a more thorough review. A financial services client adopted this model with remarkable results; their teams quickly identified the appropriate governance pathway for each AI project, eliminating the uncertainty that typically slows implementation.Related:Regular strategic reviews prove crucial. Brief, focused assessments of new AI capabilities and their business impact help catch potential issues early while identifying new opportunities. Beyond bureaucracy, it's about creating feedback loops that accelerate safe deployment while driving innovation. Teams can spot potential issues before they impact operations, transforming governance from a checkpoint into a competitive advantage.The most successful organizations have made their governance programs into strategic assets. The key question shifts from How fast can we implement AI? to How can our governance program enable faster, safer AI adoption?Start With the Business CaseA critical starting point is clear business objectives rather than technology. When teams propose AI implementations, the first question should be, What specific business process are we trying to enhance? This clarity helps build focused governance around real needs rather than hypothetical risks.Related:The enterprises succeeding with AI aren't those moving the fastest; they're moving strategically. Instead of viewing governance as a necessary burden, they should see it as a way to accelerate their AI strategy. Effective governance enables sustainable innovation that minimizes risks. In an environment where everyone feels pressured to claim they do AI the real competitive advantage comes from doing it strategically and systematically.This insight from that initial client conversation holds true: Strategic AI governance, properly designed, becomes the very engine that drives innovation forward.About the AuthorBrian BronsonChief Executive Officer, Orion InnovationBrian Bronson is CEO of Orion Innovation, a leading digital transformation and product development services firm with over 6,400 associates worldwide. A seasoned technology industry executive, Brian previously served as EVP of US Telecom, Media, and Entertainment at Capgemini, and led Radisys as President & CEO through a strategic transformation. With over 25 years of global leadership experience, Brian focuses on helping organizations maximize the impact of next-gen technologies like GenAI to drive innovation and growth.See more from Brian BronsonReportsMore ReportsNever Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.SIGN-UPYou May Also Like

Glaciers generally move so slowly you cant see their progress with the naked eye. (Their pace is glacial.) But these massive bodies of ice do march downhill, with potentially planet-altering consequences. Theres a lot we dont understand about how glaciers move and how soon some of the most significant ones could collapse into the sea. That could be a problem, since melting glaciers could lead to multiple feet of sea-level rise this century, potentially displacing millions of people who live and work along the coasts. A new group is aiming not only to further our understanding of glaciers but also to look into options to save them if things move toward a worst-case scenario, as my colleague James Temple outlined in his latest story. One idea: refreezing glaciers in place. The whole thing can sound like science fiction. But once you consider how huge the stakes are, I think it gets easier to understand why some scientists say we should at least be exploring these radical interventions. Its hard to feel very optimistic about glaciers these days. (The Thwaites Glacier in West Antarctica is often called the doomsday glaciernot alarming at all!) Take two studies published just in the last month, for example. The British Antarctic Survey released the most detailed map to date of Antarcticas bedrockthe foundation under the continents ice. With twice as many data points as before, the study revealed that more ice than we thought is resting on bedrock thats already below sea level. That means seawater can flow in and help melt ice faster, so Antarcticas ice is more vulnerable than previously estimated. Another study examined subglacial riversstreams that flow under the ice, often from subglacial lakes. The team found that the fastest-moving glaciers have a whole lot of water moving around underneath them, which speeds melting and lubricates the ice sheet so it slides faster, in turn melting even more ice. And those are just two of the most recent surveys. Look at any news site and its probably delivered the same gnarly message at some point recently: The glaciers are melting faster than previously realized. (Our site has one, too: Greenlands ice sheet is less stable than we thought, from 2016.) The new group is joining the race to better understand glaciers. Arte Glacier Initiative, a nonprofit research organization founded by scientists at MIT and Dartmouth, has already awarded its first grants to researchers looking into how glaciers melt and plans to study the possibility of reversing those fortunes, as James exclusively reported last week. Brent Minchew, one of the groups cofounders and an associate professor of geophysics at MIT, was drawn to studying glaciers because of their potential impact on sea-level rise. But over the years, I became less content with simply telling a more dramatic story about how things were goingand more open to asking the question of what can we do about it, he says. Minchew is among the researchers looking into potential plans to alter the future of glaciers. Strategies being proposed by groups around the world include building physical supports to prop them up and installing massive curtains to slow the flow of warm water that speeds melting. Another approach, which will be the focus of Arte, is called basal intervention. It basically involves drilling holes in glaciers, which would allow water flowing underneath the ice to be pumped out and refrozen, hopefully slowing them down. If you have questions about how all this would work, youre not alone. These are almost inconceivably huge engineering projects, theyd be expensive, and theyd face legal and ethical questions. Nobody really owns Antarctica, and its governed by a huge treatyhow could we possibly decide whether to move forward with these projects? Then theres the question of the potential side effects. Just look at recent news from the Arctic Ice Project, which was researching how to slow the melting of sea ice by covering it with substances designed to reflect sunlight away. (Sea ice is different from glaciers, but some of the key issues are the same.) One of the projects largest field experiments involved spreading tiny silica beads, sort of like sand, over 45,000 square feet of ice in Alaska. But after new research revealed that the materials might be disrupting food chains, the organization announced that its concluding its research and winding down operations. Cutting our emissions of greenhouse gases to stop climate change at the source would certainly be more straightforward than spreading beads on ice, or trying to stop a 74,000-square-mile glacier in its tracks. But were not doing so hot on cutting emissionsin fact, levels of carbon dioxide in the atmosphere rose faster than ever in 2024. And even if the world stopped polluting the atmosphere with planet-warming gases today, things may have already gone too far to save some of the most vulnerable glaciers. The longer I cover climate change and face the situation were in, the more I understand the impulse to at least consider every option out there, even if it sounds like science fiction. This article is from The Spark, MIT Technology Reviews weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.

Elon Musk's unique government sway could be playing a role in X's ad growth, according to EMARKETER. Graeme Sloan for The Washington Post via Getty Images 2025-03-27T12:25:48Z SaveSaved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now.Have an account? X will clock $1.3 billion in US ad revenue this year, per a new forecast from EMARKETER.The report said some of the growth is motivated by fear of Elon Musk and isn't sustainable.X's advertising tactics have garnered scrutiny from Democratic senators.Things seem to be looking up for the advertising business of Elon Musk's X.A new forecast from the research firm EMARKETER says X's advertising business will get its first annual growth since 2021 this year.EMARKETER forecasts X's US digital ad revenue will jump 17.5% to $1.3 billion this year, up from $1.1 billion in 2024. Globally, EMARKETER estimates X will pull in $2.3 billion in ad revenue this year, up 16.5% year-on-year.There's a catch, however.EMARKETER principal analyst Jasmine Enberg cautioned in her report that some of the growth is "being driven by fear" and, because of that, could be unsustainable."Many advertisers may view spending on X as a cost of doing business in order to mitigate potential legal or financial repercussions," Enberg said. "But fear is not a sustainable motivator, and the situation remains volatile, partly as some consumers' discontent toward Musk grows."Enberg's comments echo previous reporting by Business Insider.BI recently reported that ad agency execs and consultants were begrudgingly advising clients to pay what could be called an Elon tax: buying ads on X in order to avoid legal and political woes.Enberg also said some of the ad growth came from the addition of small and medium-sized businesses and that X could also stand to benefit from Meta's new lax moderation policies.X did not immediately respond to a request for comment from BI. It's a private company and doesn't publicly report ad revenue. Emarketer forecasts X's US ad revs will increase by 17.5% this year, but cautions some of the growth is being driven by fear of Elon Musk. Courtesy of Emarketer X's relationship with advertisers has been fraughtDespite the recent upswing, X's ad revenue is still much lower than it was pre-Musk. In 2022, the year Musk bought it, the company pulled in $2.4 billion in US ad revenue, per EMARKETER estimates.X's ad business plummeted in the wake of Musk's takeover. Some advertisers were wary of his changes to the company. Xlaid offa large chunk of its staff, loosened moderation, shook up account verificationrules, and brought back somebannedaccounts.X fired back at some advertisers who had spurned the platform. The company filed a lawsuit against several advertisers in August last year, accusing them of illegally conspiring to boycott the platform through their membership in a now-defunct industry initiative called the Global Alliance for Responsible Media. The case is ongoing.X's sales tactics have garnered scrutiny from Democratic senators, who sent letters to the DOJ and FTC calling for investigations. In their letters, the Senators referenced a report in The Wall Street Journal that said X's CEO, Linda Yaccarino, and a lieutenant had pushed IPG to spend more money on X, citing people with knowledge of the talks. The Journal reported that IPG execs had interpreted the message as a reminder that the Trump administration could impede its proposed $13 billion merger with the ad giant Omnicom.Disclosure: BI and EMARKETER share a parent company.

President Donald Trump's auto tariffs are set to take effect next week. Anna Moneymaker/Getty Images 2025-03-27T11:48:59Z SaveSaved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now.Have an account? Trump's auto tariffs have landed and they may have a major impact on car prices.JP Morgan analysts say 25% tariffs on cars from Mexico and Canada could push up prices by as much as $2,500.Auto stocks fell on fears the tariffs will impact the global automotive industry.After several false starts, Trump's auto tariffs are finally here and they could make buying a new car a lot more expensive.The announcement of the 25% tariffs on all imported passenger vehicles and key car parts, which take effect on April 2, sent foreign and domestic auto stocks tumbling.Analysts have long warned that the levies are likely to push up car prices as the US automotive industry scrambles to respond.Wedbush analysts said in a Wednesday note that the tariffs "would be a hurricane-like headwind to foreign (and many US) automakers and ultimately push the average price of cars up $5k to $10k depending on the make/ model/price point."Analysis from Anderson Economic Group reported by CNBC put the potential price hike from Trump's levies at between $4,000 and $12,200 depending on the vehicle. EVs are likely to be hit hard due to their imported batteries and electronics.Prior to Trump announcing a 30-day exemption for the auto industry earlier this month, JP Morgan analysts estimated that 25% tariffs on vehicle imports from Mexico and Canada alone could raise the average price of new vehicles by as much as $2,580.Automakers are also sounding the alarm. The Alliance for Automotive Innovation, which represents major US automakers including GM and Ford, warned earlier this month that proposed import taxes could see car prices rise by as much as 25%.GM and Ford, as well as foreign rivals such as Toyota, Nissan, and BMW, all have major production facilities in Mexico, meaning they are likely to be badly affected by the tariffs.Barclays analysts warned earlier this month that 25% levies on auto imports from Mexico and Canada could effectively wipe out profits for Ford, GM, and Stellantis.Automakers like Tesla that build most of their cars in the US will feel the pain too, with the tariffs extending to automobile parts including engines, powertrain parts, and "electrical components."Despite building their cars in the US, the likes of Tesla still source many of their components abroad. Between 20% and 25% of the components from Tesla's 2025 model-year vehicles come from Mexico, according to regulatory filings.Global pain Shares in foreign automakers fell following the announcement of the tariffs, with Nissan, Toyota, and Honda down between 1.5% to 3.5%.In Europe, BMW and Volkswagen fell slightly in morning trading, while Mercedes stock dipped 2% drop.Economists have previously warned that increasing the tariff of cars imported from Europe to the US to 25% from the current 2.5% would lead to a sharp fall in the number of cars shipped across the Atlantic.A January report from Oxford Economics estimated that automotive exports from Germany and Italy would decline by 7.1% and 6.6% respectively due to the tariffs. The US is the main non-EU export market for both countries.Germany's struggling auto industry would be hit particularly hard, with Oxford Economics estimating that total automotive production would decline by about 5.3%.That's not a scenario manufacturers including VW already grappling with excess capacity and mulling factory closures to cut costs will be looking forward to. VW is considering closing factories in Germany. JENS SCHLUETER / AFP US tariffs on vehicles built in China have previously led to European automakers delaying the launch of new models in the US and in the case of Chinese automakers like BYD, avoiding America entirely.The Trump administration's latest trade war escalation also raises the prospect that Europe, Japan, and South Korea may retaliate with their own tariffs on US car exports.After Trump revealed the 25% auto tariffs, Japan's prime minister vowed that "every option" was on the table as the country considered its response.Germany's economy minister, meanwhile, called on the EU to deliver a "decisive response" to the tariffs, while South Korea said it would announce emergency measures next month.

This story originally appeared in Kids Today, Voxs newsletter about kids, for everyone. Sign up here for future editions.Over the next few weeks, millions of kids around the country will come home from school, toss their backpacks in a corner, and begin an annual ritual that can be fun, relaxing, stressful, and confusing all at the same time.I speak, of course, of spring break a phrase that has historically meant one thing to beach-bound college students, and quite another to families of younger kids, facing down a week (or sometimes two) when schools are closed and work is decidedly not.Summer has its own challenges, for kids and adults alike. But spring break, like the countless other interruptions that pockmark school calendars, can be even harder to plan for. Fewer camps are open; summer school is months away. Some families go skiing or take a cruise, but amid rising prices, those options are out of reach for more and more parents. For many families, you kind of cobble it together on your own, Lauren Smith Brody, CEO of the Fifth Trimester, a workplace gender equality consultancy, told me which means a lot of stress for parents and, often, a lot of screen time for kids. Its not just an inconvenience days off of school can mean days of hunger for kids from food-insecure families, who rely on school breakfast and lunch to get through the week. And the spike in juvenile crime between the hours of 3 and 6 pm on weekdays suggests that for some kids, unsupervised time can be dangerous.Some school districts and afterschool programs offer free or low-cost spring break camps a way to reinforce some of the learning thats going on in a way that feels like the kids are having fun and having a break, as Jodi Grant, executive director of the nonprofit Afterschool Alliance, put it. But those programs were underfunded even before President Trump ordered the closure of the Education Department, which administers federal funding for afterschool programs. For now, spring break is yet another reminder of the mismatch between American work culture and the needs of human life for relaxation, for connection, for something to shake up what can sometimes be a draining daily routine. At their best, Grant said, spring break programs for kids are just a chance to do things differently.If you dont have a school-aged child, you might not realize how many days off kids have in addition to summer break. In New York City, where I live, its about two dozen, which includes winter break, mid-winter break, spring break, and a number of religious and other holidays in between.School breaks are intended, at least in part, to give families and staff time to observe said holidays (in many districts, spring break encompasses Easter, Passover, or both). Spring break, in particular, is a popular time for travel, with 48 percent of families with children planning a trip during that period, according to a 2023 TransUnion survey. But that still leaves more than half who stay put (its not that easy for one or two adults to get a full week off work at a time thats neither summer nor the winter holidays). When kids are off but parents are working, the options are somewhat limited. Some camps operate during spring break, but the cost, which can run to hundreds of dollars a week, puts them out of reach for many families. Then theres the cobble-it-together approach, with parents (and sometimes other family members like grandparents) splitting up care and work as best they can. Theres a lot of juggling, Brody said, and nobody ever gets any rest.Kids, however, do really need breaks. Take it from a student at John Jay High School in Lewisboro, New York, who wrote in the school paper in 2022 that There are so many responsibilities on high school students plate, whether they must do homework, study, work, play a sport or activity, take SAT or ACT tests, look at or visit colleges, take AP exams, etc. Having a break in the school year could help a student relax and feel okay.Experts agree that theres a reason kids dont go to class 40 hours a week, 365 days a year. Longer school days have been tried in some districts, and both kids and teachers get exhausted, Grant said. Its also really important for a lot of kids to have an environment thats not graded or judged or prescribed. So what are families supposed to do? In Philadelphia, one answer is Spring Break Camp, a free, full-day program operating at 22 schools in the district. Some publicly funded afterschool programs offer spring break and other day-off camps as well, Grant said.These programs give kids a chance to hang out with new friends and new educators, to have more choice than they typically do during the school day, and to sample activities from yoga to weightlifting to building electric cars, Grant said.Many afterschool programs also provide snacks or meals for kids. But publicly funded programs often have limited space, and private ones can be expensive. Around the country, there are nearly 25 million kids whose parents want them to be in afterschool programs, but who dont have access, largely because of cost, Grant said. Expanding access to free or low-cost afterschool programs would help families deal with the three to four hours every weekday in which parents work but schools are out, as my colleague Rachel Cohen has written. It would also provide a solution for those two dozen days every school year that leave many parents scrambling for care and many kids bored on their iPads all day long.However, federal funding for afterschool programs has not kept up with inflation, Grant said. The Trump administration and DOGE have not cut support for these programs, but that support is administered through the Education Department, which Trump has instructed Education Secretary Linda McMahon to dismantle. Were feeling the same uncertainty as everyone else, Grant said.In the absence of robust federal funding, some states, like California, are stepping up by funding afterschool programs out of their own budgets. Employers, too, can help by fostering a culture that invites people to be open about their caregiving needs, Brody said.There is, of course, a core problem at the heart of the spring break conundrum: as Brody put it, the amount of paid vacation American parents typically get (which hovers around 11-15 days for all but the most senior employees) is just so out of line with the number of weeks that there arent school.You can solve this problem with camps, or you could solve it with more paid time off. The latter, of course, feels unattainable in a time of worker precarity and a resurgent grind culture. But as kids know well, everyone deserves a break sometimes.What Im readingThe Trump administration has reportedly cut funding for a program that provides legal representation for unaccompanied immigrant children, potentially forcing them to represent themselves in immigration court.Last week, Health and Human Services Secretary Robert F. Kennedy, Jr., announced Operation Stork Speed to improve the safety of baby formula. Its not bad to devote more attention to formula, experts say but the Trump administrations cuts at the FDA could hamper safety efforts.Evereden, once a baby brand, is now trying to give Gen Alpha tweens what they want. And apparently what they want are body mists.My little kid and I have been reading Fox and His Friends, originally published in 1982 and now a time capsule of 20th-century child-rearing norms. Fox, a school-aged child/animal, is tasked with babysitting his younger sister for an entire day with no adult input whatsoever. He is terrible at his job and eventually allows her to scale a telephone pole, then bribes her for her silence with ice cream. I am not sure if this is an argument for giving kids more independence, or the opposite.From my inboxLast week, I asked for your spring break experiences. Reader Kareen H had less-than-fond memories of spring at the YMCA as a child: I NEVER wanted to be at the Y.However, Kareen did enjoy field trips, adding that I won best singer award, because I was singing to myself in the Y van, coming back from some field trip. I have no memory of what I did during spring break as a child, but I am positive I have never won a best singer award in any context. As always, thanks for your messages and get in touch any time at anna.north@vox.com.See More:

These are uncertain times for many immigrants in the US. There have been reports of individual visa and green card holders and tourists who have been detained and deported. However, the Trump administration does not seem to be indiscriminately targeting legal immigrants who have authorization to be in the US on a large scale. Some have reportedly been targeted based on their political activism. A Brown University professor and doctor with a green card was deported after officials found photos of former Hezbollah leader Hassan Nasrallah and Irans supreme leader on her phone. Immigration authorities also invoked a Trump executive order prohibiting antisemitism to detain a former Columbia student and green card holder who helped lead campus protests over the war in Gaza.In other cases, the Trump administration hasnt clarified its rationale for detaining someone. A German citizen with a green card was interrogated by border officials in Boston and detained without access to his anxiety medication. Its not clear if the government has charged him with a crime. Similarly, the administration had not offered an official explanation for detaining a Turkish doctoral student as of Tuesday.And its not just immigrants who have been affected. A US citizen said he was walking down the streets of Chicago when he was arrested by immigration agents, who confiscated his ID and held him for 10 hours before releasing him. Even though limited in number, these cases have been going viral and are understandably causing fear in immigrant communities.According to immigration attorneys, its hard to tell how worried immigrants who are legally living and working in the US need to be. After practicing for 40 years, its really difficult to divine what a measured response is right now, said Kathleen Campbell Walker, former president of the American Immigration Lawyers Association. So far, cases of individuals with visas and green cards being detained or deported appear rare. That said, the lawyers I spoke with advised these immigrants, as well as US citizens, to consider certain precautions in an environment of such uncertainty.Consider carrying identification documentsLegal non-citizen immigrants have long been legally required to carry their immigration papers at all times. However, the penalties for failing to do so are becoming higher under Trump.In April, the Trump administration is expected to increase the associated fines from $100 to $5,000, Campbell Walker said. Failing to possess documentation is a misdemeanor. That could now land an immigrant in detention and deportation proceedings; Trump revoked the Biden administrations immigration enforcement priorities so that even people charged with nonviolent, minor crimes can now be deported.Relatedly, next month the Trump administration will also start requiring all noncitizens to register with the federal government and designate those who fail to do so as a priority for immigration enforcement. Many noncitizens who have had previous contact with the federal government whether because they applied for certain immigration benefits or were issued a notice to appear in immigration court are already considered registered under the new policy.Campbell Walker said US citizens should also consider carrying a passport card that fits in their wallet, or birth certificate, as proof of their nationality, given the reports of Americans swept up in Trumps immigration enforcement activities. In some of these cases, she said, there have been concerns that immigration agents are racially and ethnically profiling their targets.Carrying documents on your person, making sure that people who are not citizens or naturalized or acquired citizens have one place in your home where you have all your important documentation together and making sure that you have copies those are all reasonable and important steps to be taking in a moment like this, when we see the administration attacking free speech rights and attacking the basic norms of due process, said Heidi Altman, vice president of policy at the National Immigration Law Center.Reconsider international travelImmigration attorneys are urging immigrants to exercise caution in traveling abroad right now.Following the deportation of one of their professors, Brown University advised out of an abundance of caution that even green card holders delay any personal travel outside the US. The university said that changes in requirements to reenter the country and a draft proposal for a travel ban targeting 43 countries that could be implemented as early as this week might impact its students and staff. I believe that a lot of green card holders are making the decision to consult with an attorney before traveling, and I think thats a reasonable consideration, Altman said. Immigrants should consider whether their country of origin or where they are planning to travel may be on the list of countries that could be subjected to travel bans. They should also weigh their own history of activism and whether that could make them a target upon reentry to the US. We know that this administration is engaging in retaliatory actions against people who have engaged in constitutionally protected activism and speech, Altman said. And so I think people may want to think about their own history and imagine and explore if it might put them at high interest for retaliatory targeting and talk to an attorney about precautionary steps that can be taken before travel.If you must travel, consider leaving your personal electronic devices at home. Border officials can (and recently have) requested access to immigrants devices, including their cellphones. Refusing to grant them access might give them grounds to deny entry on the basis that they have insufficient information to determine if an immigrant is admissible to the US. But Campbell Walker said that she is concerned about officers lacking the training required to appropriately evaluate whats on a personal device. She said that, based on reports from member attorneys of the American Immigration Lawyers Association, officials are now looking through social media feeds on peoples phones for reasons to deny them admission to the US.Im not asking anyone to lie. Im not trying to obstruct justice, she said. But if somebody who may not have sufficient training is going to rip through a cellphone and jump to conclusions and potentially remove me or prevent me from entering the US, I dont think its advisable to have a bunch of social media or photographs on the phone you travel with. I dont think its very wise to be traveling with your [personal] laptop.See More:

It is now more or less impossible to put a precise figure on the number of video games released each year. According to data published by the digital store Steam, almost 19,000 titles were released in 2024 and thats just on one platform. Hundreds more arrived on consoles and smartphones. In some ways this is the positive sign of a vibrant industry, but how on earth does a new project get noticed? When Triple A titles with multimillion dollar marketing budgets are finding it hard to gain attention (disappointing sales have been reported for Dragon Age: The Veilguard, the Final Fantasy VII remakes and EA Sports FC), what chance is there for a small team to break out?And yet it does happen. Last years surprise hit Balatro has shifted more than 5m copies. Complex medieval strategy title Manor Lords sold 1m copies during its launch weekend. But what awaits a small developer after they achieve success? And what does success even mean in a continuously evolving industry?James Carbutt and Will Todd of Coal Supper are still trying to make sense of it. Their acerbic satire Thank Goodness Youre Here!, in which players slap their way through bizarre quotidian scenarios in the fictional humble northern town of Barnsworth, is now an award-winning game. Its just not registered as a success in my head at all, says Carbutt. The numbers are going up on screen, and there have been YouTube playthroughs and some erotic fan art. Beyond that, it wont register.I cant imagine making any more games. I dont know where I would go from hereAfter spending three years working on the project, the pair now find themselves in the confusing glare of the spotlight, fielding questions about whats next. Its horrible, Carbutt jokes. But I dont think we feel any sort of second-album syndrome. The space it gives you to be a bit introspective about what you want to do next is the interesting quirk of a successful indie game.Veteran indie developer Gabe Cuzzillo (Ape Out, Baby Steps) offered them sage wisdom. He spoke about how you should focus not just on making something good because how do you quantify that, its amorphous? says Todd. Instead we should look at what it is we want to explore and judge success intrinsically, based on whether we explored that thing. The pressure of speed to market doesnt apply to us, because its never going to be possible to crank something out in six months to chase success anyway. Its more like, in the wake of this being received well, whats the next thing we want to explore? Thats something were interrogating at the moment.Australian developer Grace Bruxner has also redefined success after leaving behind a trilogy of Frog Detective games: bite-size adventures co-developed with Thomas Bowker that quickly became cult indie hits.Has it impacted peoples lives in a positive way? Frog Detective. Photograph: WormclubSuccess in games has always been a bit of a lie, a bit of an illusion, she says, pointing to typical markers such as cultural impact, player numbers and financial gain. My measure of success is: did I make something Im proud of, and has it impacted my life and other peoples lives in a positive way? And yes, it did, so thumbs up.Bruxner began working on the series during her final year at university as an experiment, to see whether she could produce a commercial game. After a relatively breezy first outing, the second Frog Detective game demanded that Bruxner and Bowker lock in, and spend most of their time on the project. By the third instalment, the hard work had paid off, though the pressure had begun to take its toll. Throw in the pandemic, as well as mental and physical health issues, and Bruxner was ready to take a break. I wasnt grinding super hard, but I also wasnt having a great time, she says. It just was really nice to make that choice to stop.skip past newsletter promotionSign up to Pushing ButtonsFree weekly newsletterKeza MacDonald's weekly look at the world of gamingPrivacy Notice: Newsletters may contain info about charities, online ads, and content funded by outside parties. For more information see our Privacy Policy. We use Google reCaptcha to protect our website and the Google Privacy Policy and Terms of Service apply.after newsletter promotionIm looking forward to making art again, instead of feeling like I have to show up to a job I never applied toBruxner still has game ideas swirling in the back of her head, but she wanted to escape the endless production cycle that has swallowed up many of her peers, regardless of mounting exhaustion or burnout. Its not universal advice, she says, but if youre a solo dev or a really small team, I dont think theres any shame in leaving it there. Unless you love making games. Im not sure I love making games. I was quite young when we released the first Frog Detective, so it was like, This is my entire identity for life. I dont know how to be a separate person from that.Three years after the series swan song, she is on indefinite hiatus, exploring alternative creative paths such as pottery. I cant imagine making games, because of the expectations on me as a creator, she explains. I dont even know where I would go from here.Bruxner has been surprised by her ability to sustain herself on the modest amount of money provided by Frog Detective. If your game continues to have a tail, and you can budget properly and live within your means, it is possible to have a passive income that isnt tied to being a horrible landlord, she explains. Even so, she knows how taboo it can be to talk plainly about money, especially in creative circles like the indie game scene. I have the free time to chill and decide what I want to do, but I assume at some point Ill probably need to have a career again. My biggest question is will this money last forever? Probably not, and then what happens when it runs out? I dont know.Opportunities are limited Consume Me. Photograph: HexecutableIt may seem as though more indies than ever have broken into the spotlight in recent years. But enduring games industry turbulence has made finding financial support for follow-ups and debuts more complicated. The elephant in the room is everything thats happened over the past couple of years, with mass layoffs, studio closures and evaporating funding opportunities, explains AP Thomson, a developer of the forthcoming indie Consume Me with fellow NYU Game Center graduate Jenny Jiao Hsia. Before that, there was a pretty major change around the mid-2010s when indie publishers and funders started rising in prominence. Everything weve heard suggests that the same opportunities no longer exist or are incredibly limited.Consume Me, the duos coming-of-age scheduling RPG doesnt have a release date but has already been nominated for five gongs at the Independent Games Festival awards. As such, Jiao Hsia and Thomson are already under pressure to decide their next endeavour. Multiple people have told us we should be moving forward once it launches, says Thomson.Even with growing expectations, the pair arent keen to get ahead of themselves. Everything weve heard suggests that now is really not a great time to be pitching, so were going to focus our energy on the launch and then read the temperature of the room after that, Thomson adds.Im looking forward to finding enjoyment in making art again, instead of feeling like I have to show up to a job I never applied to, explains Jiao Hsia. The idea of making art for fun, without worrying about making money off it, is something I cant wait to do.