The popular virtual meeting service has been hit with a major outage. Time to call it a day?

Why create an AI model focused solely for the enterprise? Answer by Sean White.

In context: Prompt injection is an inherent flaw in large language models, allowing attackers to hijack AI behavior by embedding malicious commands in the input text. Most defenses rely on internal guardrails, but attackers regularly find ways around them – making existing solutions temporary at best. Now, Google thinks it may have found a permanent fix. Since chatbots went mainstream in 2022, a security flaw known as prompt injection has plagued artificial intelligence developers. The problem is simple: language models like ChatGPT can't distinguish between user instructions and hidden commands buried inside the text they're processing. The models assume all entered (or fetched) text is trusted and treat it as such, which allows bad actors to insert malicious instructions into their query. This issue is even more serious now that companies are embedding these AIs into our email clients and other software that might contain sensitive information. Google's DeepMind has developed a radically different approach called CaMeL (Capabilities for Machine Learning). Instead of asking artificial intelligence to self-police – which has proven unreliable – CaMeL treats large language models (LLMs) as untrusted components inside a secure system. It creates strict boundaries between user requests, untrusted content like emails or web pages, and the actions an AI assistant is allowed to take. CaMeL builds on decades of proven software security principles, including access control, data flow tracking, and the principle of least privilege. Instead of relying on AI to catch every malicious instruction, it limits what the system can do with the information it processes. Here's how it works. CaMeL uses two separate language models: a "privileged" one (P-LLM) that plans actions like sending emails, and a "quarantined" one (Q-LLM) that only reads and parses untrusted content. The P-LLM can't see raw emails or documents – it just receives structured data, like "email = get_last_email()." The Q-LLM, meanwhile, lacks access to tools or memory, so even if an attacker tricks it, it can't take any action. All actions use code – specifically a stripped-down version of Python – and run in a secure interpreter. This interpreter traces the origin of each piece of data, tracking whether it came from untrusted content. If it detects that a necessary action involves a potentially sensitive variable, such as sending a message, it can block the action or request user confirmation. Simon Willison, the developer who coined the term "prompt injection" in 2022, praised CaMeL as "the first credible mitigation" that doesn't rely on more artificial intelligence but instead borrows lessons from traditional security engineering. He noted that most current models remain vulnerable because they combine user prompts and untrusted inputs in the same short-term memory or context window. That design treats all text equally – even if it contains malicious instructions. // Related Stories CaMeL still isn't perfect. It requires developers to write and manage security policies, and frequent confirmation prompts could frustrate users. However, in early testing, it performed well against real-world attack scenarios. It may also help defend against insider threats and malicious tools by blocking unauthorized access to sensitive data or commands. If you love reading the undistilled technical details, DeepMind published its lengthy research on Cornell's arXiv academic repository.

Shopping around for a pair of noise-canceling headphones can be a little daunting. One recommendation we would like to make is to try and stick with a brand you already own some devices from. As a matter of fact, if you own one or more Apple devices, you’re an excellent candidate for the Apple AirPods Max.  These premium ANC headphones received a USB-C overhaul in 2024, and today these Apple cans are discounted to $480. The full MSRP on these over-ears is $550.  First and foremost, if you at Related The AirPods Max use an Adaptive EQ that automatically optimizes sound quality based on fit and your listening environment. Out of the box, the AirPods Max have solid midrange articulation and a good amount of low-end that makes rock, rap, and hip-hop tracks extra fun to listen to.  The headphones also have an adaptive ANC system that puts a damper on distracting noise, including vehicle engines and workplace chatter. On a full charge, you should get up to 20 hours with ANC enabled, and just 5 minutes of charging should net you an extra 1.5 hours of battery life.  While there’s no companion app for changing audio presets or adjusting ANC controls, iOS users can access several customizations under Bluetooth and Accessibility.  Save $70 when you purchase the Apple AirPods Max today, and be sure to take a look at our lists of the best AirPods deals, best headphone deals, and best Apple deals for even more discounts on top Apple products. Editors’ Recommendations

A momentous exhibition at the Prado Museum reunites all but one of the canvases El Greco painted for the Spanish church where he was buried.

simulacrum of thought OpenAI releases new simulated reasoning models with full tool access New o3 model appears "near-genius level," according to one doctor, but it still makes mistakes. Benj Edwards – Apr 16, 2025 6:21 pm | 3 Credit: Floriana via Getty Images Credit: Floriana via Getty Images Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more On Wednesday, OpenAI announced the release of two new models—o3 and o4-mini—that combine simulated reasoning capabilities with access to functions like web browsing and coding. These models mark the first time OpenAI's reasoning-focused models can use every ChatGPT tool simultaneously, including visual analysis and image generation. OpenAI announced o3 in December, and until now, only less capable derivative models named "o3-mini" and "03-mini-high" have been available. However, the new models replace their predecessors—o1 and o3-mini. OpenAI is rolling out access today for ChatGPT Plus, Pro, and Team users, with Enterprise and Edu customers gaining access next week. Free users can try o4-mini by selecting the "Think" option before submitting queries. OpenAI CEO Sam Altman tweeted that "we expect to release o3-pro to the pro tier in a few weeks." For developers, both models are available starting today through the Chat Completions API and Responses API, though some organizations will need verification for access. "These are the smartest models we've released to date, representing a step change in ChatGPT's capabilities for everyone from curious users to advanced researchers," OpenAI claimed on its website. OpenAI says the models offer better cost efficiency than their predecessors, and each comes with a different intended use case: o3 targets complex analysis, while o4-mini, being a smaller version of its next-gen SR model "o4" (not yet released), optimizes for speed and cost-efficiency. OpenAI says o3 and o4-mini are multimodal, featuring the ability to "think with images." Credit: OpenAI What sets these new models apart from OpenAI's other models (like GPT-4o and GPT-4.5) is their simulated reasoning capability, which uses a simulated step-by-step "thinking" process to solve problems. Additionally, the new models dynamically determine when and how to deploy aids to solve multistep problems. For example, when asked about future energy usage in California, the models can autonomously search for utility data, write Python code to build forecasts, generate visualizing graphs, and explain key factors behind predictions—all within a single query. OpenAI touts the new models' multimodal ability to incorporate images directly into their simulated reasoning process—not just analyzing visual inputs but actively "thinking with" them. This capability allows the models to interpret whiteboards, textbook diagrams, and hand-drawn sketches, even when images are blurry or of low quality. That said, the new releases continue OpenAI's tradition of selecting confusing product names that don't tell users much about each model's relative capabilities—for example, o3 is more powerful than o4-mini despite including a lower number. Then there's potential confusion with the firm's non-reasoning AI models. As Ars Technica contributor Timothy B. Lee noted today on X, "It's an amazing branding decision to have a model called GPT-4o and another one called o4." Vibes and benchmarks All that aside, we know what you're thinking: What about the vibes? While we have not used 03 or o4-mini yet, frequent AI commentator and Wharton professor Ethan Mollick compared o3 favorably to Google's Gemini 2.5 Pro on Bluesky. "After using them both, I think that Gemini 2.5 & o3 are in a similar sort of range (with the important caveat that more testing is needed for agentic capabilities)," he wrote. "Each has its own quirks & you will likely prefer one to another, but there is a gap between them & other models." During the livestream announcement for o3 and o4-mini today, OpenAI President Greg Brockman boldly claimed: "These are the first models where top scientists tell us they produce legitimately good and useful novel ideas." Early user feedback seems to support this assertion, although until more third-party testing takes place, it's wise to be skeptical of the claims. On X, immunologist Dr. Derya Unutmaz said o3 appeared "at or near genius level" and wrote, "It's generating complex incredibly insightful and based scientific hypotheses on demand! When I throw challenging clinical or medical questions at o3, its responses sound like they're coming directly from a top subspecialist physicians." OpenAI benchmark results for o3 and o4-mini SR models. Credit: OpenAI So the vibes seem on target, but what about numerical benchmarks? Here's an interesting one: OpenAI reports that o3 makes "20 percent fewer major errors" than o1 on difficult tasks, with particular strengths in programming, business consulting, and "creative ideation." The company also reported state-of-the-art performance on several metrics. On the American Invitational Mathematics Examination (AIME) 2025, o4-mini achieved 92.7 percent accuracy. For programming tasks, o3 reached 69.1 percent accuracy on SWE-Bench Verified, a popular programming benchmark. The models also reportedly showed strong results on visual reasoning benchmarks, with o3 scoring 82.9 percent on MMMU (massive multi-disciplinary multimodal understanding), a college-level visual problem-solving test. OpenAI benchmark results for o3 and o4-mini SR models. Credit: OpenAI However, these benchmarks provided by OpenAI lack independent verification. One early evaluation of a pre-release o3 model by independent AI research lab Transluce found that the model exhibited recurring types of confabulations, such as claiming to run code locally or providing hardware specifications, and hypothesized this could be due to the model lacking access to its own reasoning processes from previous conversational turns. "It seems that despite being incredibly powerful at solving math and coding tasks, o3 is not by default truthful about its capabilities," wrote Transluce in a tweet. Also, some evaluations from OpenAI include footnotes about methodology that bear consideration. For a "Humanity's Last Exam" benchmark result that measures expert-level knowledge across subjects (o3 scored 20.32 with no tools, but 24.90 with browsing and tools), OpenAI notes that browsing-enabled models could potentially find answers online. The company reports implementing domain blocks and monitoring to prevent what it calls "cheating" during evaluations. Even though early results seem promising overall, experts or academics who might try to rely on SR models for rigorous research should take the time to exhaustively determine whether the AI model actually produced an accurate result instead of assuming it is correct. And if you're operating the models outside your domain of knowledge, be careful accepting any results as accurate without independent verification. Pricing For ChatGPT subscribers, access to o3 and o4-mini is included with the subscription. On the API side (for developers who integrate the models into their apps), OpenAI has set o3's pricing at $10 per million input tokens and $40 per million output tokens, with a discounted rate of $2.50 per million for cached inputs. This represents a significant reduction from o1's pricing structure of $15/$60 per million input/output tokens—effectively a 33 percent price cut while delivering what OpenAI claims is improved performance. The more economical o4-mini costs $1.10 per million input tokens and $4.40 per million output tokens, with cached inputs priced at $0.275 per million tokens. This maintains the same pricing structure as its predecessor o3-mini, suggesting OpenAI is delivering improved capabilities without raising costs for its smaller reasoning model. Codex CLI OpenAI also introduced an experimental terminal application called Codex CLI, described as "a lightweight coding agent you can run from your terminal." The open source tool connects the models to users' computers and local code. Alongside this release, the company announced a $1 million grant program offering API credits for projects using Codex CLI. A screenshot of OpenAI's new Codex CLI tool in action, taken from GitHub. Credit: OpenAI Codex CLI somewhat resembles Claude Code, an agent launched with Claude 3.7 Sonnet in February. Both are terminal-based coding assistants that operate directly from a console and can interact with local codebases. While Codex CLI connects OpenAI's models to users' computers and local code repositories, Claude Code was Anthropic's first venture into agentic tools, allowing Claude to search through codebases, edit files, write and run tests, and execute command line operations. Codex CLI is one more step toward OpenAI's goal of making autonomous agents that can execute multistep complex tasks on behalf of users. Let's hope all the vibe coding it produces isn't used in high-stakes applications without detailed human oversight. Benj Edwards Senior AI Reporter Benj Edwards Senior 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. 3 Comments

Health It is harder for our bodies to absorb key nutrients from plant-based foods, so some vegans may be short on essential amino acids for healthy muscles and bones despite eating plenty of protein 16 April 2025 It isn’t necessarily easy to absorb important amino acids from vegan foodsSolStock/Getty Images While most vegans eat enough protein, it seems many fall short on essential amino acids such as lysine and leucine, according to a new study. The finding suggests vegans may need to be more careful about the kind of protein in their diet, not just the quantity they consume. Patricia Soh at the Riddet Institute in Palmerston North, New Zealand, and her colleagues analysed the diets of 193 adults living in New Zealand who had been vegan for at least two years prior to the study.…

The popular UPPABaby stroller will go from $899 to $1199 due to tariffs. Fly View Productions/Getty Images 2025-04-16T22:07:27Z Save Saved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now. Have an account? UPPAbaby brand strollers and car seats are going to cost parents more because of tariffs, one store owner tells me. Elizabeth Mahon, the owner of a baby goods store in Washington, DC, is raising prices on one stroller by $300. UPPAbaby says it's trying to absorb increased costs of production, but calls price hikes "unavoidable." Parents, get ready: It looks like it's going to get a lot more expensive to raise a family — and some places are already raising prices, like one baby and toy store in Washington, DC.I talked to Elizabeth Mahon, the owner of Three Littles, which she's run since 2019. She told me price hikes are on the way, and she said they're because of Donald Trump's tariffs.She said she'd have to raise prices on items that are manufactured mostly in China, like the popular UPPAbaby Vista stroller, which will now cost $1,199 — up from $899.Massachusetts-based UPPAbaby has said it's passing the increased costs of production onto customers. "Please know that we've made every effort behind the scenes to absorb as much of the cost as possible," the company said in a statement on its Instagram, "but some price increases are unfortunately unavoidable."Mahon told me how tariffs are already affecting her store and her customers. And about her biggest worry: that some families won't be able to afford necessities like car seats at all.(This interview has been edited for length and clarity).Business Insider: So what's going on with the price hikes with strollers and car seats?Elizabeth Mahon: Nuna and UPPAbaby both reached out about a week ago and said, "Hey, this is going to be a new price list moving forward."Then, on Monday, we got another email from UPPAbaby with an amended price list. When they first emailed, it was when tariffs were at 54% and now tariffs are 145% — they had to fix the prices to better reflect the current tariffs.Is this across the whole range of things? Strollers, car seats, accessories? Yes, pretty much everything is going up in price. Initially, a lot fewer items were included in these price increases. There were still price increases, but they weren't as drastic or far-reaching. And now, the prices of just about everything that they manufacture have to go up.So the UPPABaby Vista stroller, which is about $900 for the base model — that's going up about $300. It's basically like a 33% increase. Is that about what you're seeing across the board? 33%?It depends. Some things are going up 20%, some things are going up 40%. It just depends. But 30% is, I would say, the average.Have you heard from other manufacturers about tariff increases in other categories? Clothing, toys, books?A ton. I've been getting a phone call or an email a day from our vendors letting us know that the prices are increasing. We've received a couple of emails from vendors telling us that they will no longer even be able to entertain a wholesale program.I do know that some other stroller and car seat brands are chatting about what it would look like to sell exclusively online. This would be a huge detriment to families because they won't be able to test them out in person.What are you hearing from your customers?I think there's a lot of panic buying. People are hurrying up to buy car seats that maybe they don't need yet because they are worried about these price increases. We've been talking to a lot of customers who are considering buying for children they don't even have yet so that they can get these prices.I would be remiss not to acknowledge that we have a shop in a community where there's a lot of privilege.I've heard a lot on social media that these price hikes aren't going to deter people who are shopping for these more expensive brands. But I absolutely disagree. I've heard daily from people who have said, "This stroller at $899 was already a huge splurge for me, and $1,200 is just too far out of my price range."Beyond that, the bigger conversation I think is that there are going to be many families who just cannot afford a stroller for their kids or a car seat for their child, period. Even the lower-end products — there is no option to just buy exclusively American.There is no option to just not use a car seat. You have to have it, but we're going to see price increases across every brand no matter what if they're manufactured in China.With car seats, that's the kind of item that — unlike strollers — it's really recommended you don't get one used, right?Absolutely. I am a child passenger safety technician, and we just would never recommend getting a secondhand car seat because there is no way to prove that that car seat has not been in a crash, which makes it no longer safe to use.As a small-business owner, how will these price increases affect your bottom line?I've just had to make some really big decisions already and spend money that we didn't really have to spend on extra inventory because these tariff hikes and the hysteria surrounding these new inflated prices are causing a lot of people to buy.There are going to be a lot of gaps where people aren't going to be able to buy things when they need them. That's because people are buying convertible car seats today for kids who won't fit in it for a year. The people who need that car seat now might not be able to get it because it's sold out. The manufacturing facilities can't produce at a faster rate than they had already planned on, or they don't have the materials, or they don't have the money, or they're trying to navigate the tariffs.Are you worried about what will happen having to pass on the prices to customers if that will turn customers away from your store?Absolutely. We already have customers who will come into the shop and demo the products we have and then come back in and tell us they bought them secondhand. And that's everyone's prerogative. I totally understand why people do that. The baby industry is expensive and the products we sell are on the higher-end. But I think that we'll be starting to see a lot more of that.I do think we'll see the resale market inching up higher, which will make some of these products that people used to be able to buy secondhand unaffordable, too.Even more than business, I'm worried that there are going to be families that aren't able to get seats for their children at all because they can't afford them. Recommended video

This story appeared in The Logoff, a daily newsletter that helps you stay informed about the Trump administration without letting political news take over your life. Subscribe here.Welcome to The Logoff: Today I’m focusing on two federal judges threatening to hold the Trump administration in contempt, escalating a conflict between the president and the courts with major implications for the balance of power in our republic.What’s the latest? Judge James Boasberg said today that, if President Donald Trump’s team does not give the dozens of Venezuelan men sent to a Salvadorian prison a chance to legally challenge their removal, he’ll begin contempt proceedings against the administration. In March, Boasberg ordered the administration to halt the deportation flights — and to turn around any planes that had already taken off. The administration did not comply.In a separate case, Judge Paula Xinis demanded Tuesday that the administration answer questions about why it was not complying with a Supreme Court order to “facilitate” the release of a man who was — by the administration’s own admission — sent to a Salvadorian prison via an “administrative error.” Kilmar Armando Abrego Garcia’s lawyers are asking the court to hold the administration in contempt, but Xinis said she first wanted to review its responses.What’s next? A Trump spokesperson said today that the administration would appeal Boasberg’s ruling. In the Xinis case, the administration has two weeks to comply with the judge’s request for answers.What happens if a judge finds the administration in contempt of court? The judicial branch largely relies on the executive branch to enforce its decisions, including imposing consequences for contempt of court. But what happens when you ask the executive branch — and particularly this executive branch — to impose consequences on itself? That’s not clear.What’s the big picture? Do judges have the power to compel the administration to change its behavior? The results of these two cases will go a long way toward providing an answer to that crucial question. And if that answer is “no,” then the single most powerful check on Trump’s power will be greatly diminished.And with that, it’s time to log off…I really enjoyed this interview with Tina Fey, both because she’s a tremendously entertaining and innovative thinker and because she’s a living reminder that we can find joy in difficult times. Thanks so much for reading, and I’ll see you back here tomorrow.You’ve read 1 article in the last monthHere at Vox, we're unwavering in our commitment to covering the issues that matter most to you — threats to democracy, immigration, reproductive rights, the environment, and the rising polarization across this country.Our mission is to provide clear, accessible journalism that empowers you to stay informed and engaged in shaping our world. By becoming a Vox Member, you directly strengthen our ability to deliver in-depth, independent reporting that drives meaningful change.We rely on readers like you — join us.Swati SharmaVox Editor-in-ChiefSee More:

The just-ended Daredevil: Born Again focuses on Matt Murdock’s ongoing clash with the Kingpin, but it’s another major rogue that sets things in motion. Minutes into the pilot, Wilson Bethel’s Bullseye shows up, kills Foggy with a single gunshot, and proceeds to wreak havoc inside of Josie’s Bar. When he’s not throwing a lot of knives at Daredevil, he’s throwing knives into civilians, knocking them out with cue balls to the skull, or just punching them in the stomach. Many of these could’ve been easily bypassed while still keeping Daredevil on his toes, but among his many skills and traits, ol’ Bullseye is vindictive and such a hater. To be a Marvel fan is to have mixed feelings about the MCU’s villain output. For every Killmonger or Vulture, you’ve got a Cassandra Nova, Iron Monger, or whatever Kang was teed up to be. After a certain point, the baddies stopped being simple foils for the good guys and became saddled with attempts at being topical or thematically relevant in some respect. This can work, but too often it’s left projects feeling scattered or confused. For many, the most recent example will be Captain America: Brave New World’s Leader and Red Hulk, but there’s also the Skrulls, Thanos, and Falcon & the Winter Soldier’s double dose of John Walker and Karli Morgenthau. This is a problem that goes beyond a single franchise—it might extend to all blockbuster action movies in the last 20 years—but Marvel’s been the worst at it for sure. And yet, the current slate of MCU projects has featured some memorable haters among their ranks. Much of this has come from TV: both Daredevil shows have avoided this in part by just letting Bullseye, Kingpin, and Fisk’s wife Vanessa be some of the most evil, mean people in all of New York. All three are victims to one degree or another, but both the original Daredevil and Born Again have stopped short of making them overly sympathetic or trying to distance them from their previously bad actions. The Fisk who loves Vanessa like she’s the only woman in the world is the same man who locked her lover in a cage, decapitated a man with a car door, and crushed another man’s head with his bare hands. Vanessa, who loves her husband just as much, viewed his years-long departure as a betrayal on par with her father cheating on her mother—yet she still orchestrated Foggy’s death and shot her lover dead like it was nothing. © Marvel Studios The same is true of Ben Poindexter, aka Bullseye, maybe the most broken person in Daredevil’s assortment of characters. Yes, he’s forever wedded to those two as they deploy him like an attack dog, which sometimes ends with him getting thrown off a building or his spine broken. But he likes what he does and has fun messing with Matt, either by wearing his costume or throwing anything in the room at him when they fight—how can the guy not get a kick out of this when he can kill a fly with a paper clip or toss a gun at someone’s head without looking? These contrasts help make these characters compelling while also keeping them true to themselves as they test Matt’s resolve and patience. Outside of Born Again, Marvel’s found another successful hater villain in Your Friendly Neighborhood Spider-Man’s take on Scorpion. For whatever reason—maybe he’s not as visually dynamic as other heavy hitters—he hasn’t been given his own spotlight like Spider-Man’s A-list baddies Doctor Octopus, Green Goblin, or Venom. Over the past decade, Mac Gargan has come part of a larger set; Insomniac’s Spider-Man games made him part of the Sinister Six (and later paired him with Rhino), and his one live-action appearance in Spider-Man: Homecoming seemed ready to continue that trend in a hypothetical sequel. Friendly Neighborhood’s solution? Individualize the two most well-known incarnations but make them equally formidable. Gargan is introduced in the show as the leader of an emerging gang called the Scorpions making trouble for the One-Tenth, and his eventual successor Carmilla Black as his second-in-command. There’s not much nuance to them, but they make up for it in presence. Carmilla’s introduced as a problem Lonnie and the One-Tenth will have to deal with eventually, but she pales in comparison to Gargan, who’s not just a hater, but a complete demon. A scary guy is even scarier in an armored suit with a stinger tail, and he kills and terrorizes the city like he’s finally found his calling in life. In Gargan, the show’s creative team channels the same energy Homecoming did with its shovel talk scene between Peter and Vulture. Friendly Neighborhood seems to have a bright future ahead of it, and with luck, Gargan and Black will pop up every now and then to put Peter and any other potential heroes through their paces. Now all we need is for that energy to extend over to the movies. Captain America: Brave New World took some steps in that direction with Giancarlo Esposito’s Sidewinder, who just wants to kill Sam, whether he gets paid or not. Their beef and subsequent brawls are enough to make you wish that was just the whole movie, and speaks to what makes hero/villain rivalries engaging when done well. Such a personal, ongoing fight is more interesting than a villain who thinks they’re a hero or working toward some grand plan set to pay off years later, and it’s something the MCU’s conveyor belt of projects could stand to have more of, regardless of the medium. Want more io9 news? Check out when to expect the latest Marvel, Star Wars, and Star Trek releases, what’s next for the DC Universe on film and TV, and everything you need to know about the future of Doctor Who.