Author Archives: Steve Dent

Meta may trade AI chips for shares in its latest AMD deal

Posted on by

Meta has struck a deal with AMD to buy up to six gigawatts worth of AI chips, both companies announced. The agreement is structured in a way that could see AMD issue Meta up to 160 million shares of its common stock provided GPU shipment milestones are achieved — meaning Meta could own up to 10 percent of AMD if the deal fully completes. 

Meta plans to purchase six gigawatts of AMD's Instinct GPUs based on the MI450 architecture and optimized for Meta's workloads, with the first gigawatt deployment set to begin in the second half of 2026. AMD and Meta will also expand on their EPYC CPU partnership, with Meta deploying "millions" of AMD EPYC CPUs and become a launch customer for its sixth-generation EPYC CPUs. 

The tranche of AMD common stock will vest with the first one gigawatt of shipments, with additional tranches vesting as Meta scales to 6 gigawatts. Vesting is tied to AMD hitting certain stock price thresholds and Meta achieving certain technical and commercial milestones. The deal is very similar to one that AMD structured with OpenAI last year, with AMD obtaining up to a 10 percent stake in AMD in exchange for six gigawatts of Instinct GPUs. 

Such deals are being likened to circular transactions that have created a tangle of interconnected dependencies between AI companies and chip manufacturers. Analysts have observed that such deals may magnify losses if demand for AI doesn't match the sky-high market expectations. 

The agreement also shows that AI companies are keen to diversify away from NVIDIA, with AMD being a key alternative. "By diversifying our partnerships and technology stack, we’re building a more resilient and flexible infrastructure," Meta wrote in its news release. 

This article originally appeared on Engadget at https://www.engadget.com/ai/meta-may-trade-ai-chips-for-shares-in-its-latest-amd-deal-134259842.html?src=rss

The US military will reportedly use Elon Musk’s Grok AI in its classified systems

Posted on by

The US Department of Defense has reportedly reached a deal to use Elon Musk's Grok in its classified systems, according to Axios. That follows news that the Pentagon is currently in a dispute with another AI company, Anthropic, over limits on its technology for things like mass surveillance.

Last year, the White ordered Grok, along with ChatGPT, Gemini and Anthropic's Claude to be approved for government use. Up until now, though, only Anthropic's model has been allowed for the military's most sensitive tasks in intelligence, weapons development and battlefield operations. Claude was reportedly used in the Venezuelan raid in which the US military exfiltrated the country's president, Nicolás Maduro, and his wife. 

However, the Pentagon demanded that Anthropic make Claude available for "all lawful purposes" including mass surveillance and the development of fully autonomous weapons. Anthropic reportedly refused to offer its tech for those things, even with a "safety stack" built into that model. 

xAI, by contrast, agreed to a standard that would allow the DoD to employ its AI for any purpose it deems "lawful." However, the xAI model is not considered by officials to be as cutting-edge or reliable as Anthropic's Claude, and they admit that replacing Claude with Grok would be a challenge. The Pentagon is reportedly also negotiating deals with OpenAI and Gemini, both of which it considers to be on par with Anthropic.

xAI had announced a version of Grok for US government agencies in July 2025. Shortly before that, though, the chatbot started spouting fascist propaganda and antisemitic rhetoric while dubbing itself "MechaHitler." All of that followed a public spat between Musk and Trump over the president's spending bill, after which GSA approval of Grok seemed to stall. Earlier this week, Anthropic accused three Chinese AI labs of abusing Claude's AI with "distillation attacks" to improve their own models. 

This article originally appeared on Engadget at https://www.engadget.com/ai/the-us-military-will-reportedly-use-elon-musks-grok-ai-in-its-classified-systems-110049021.html?src=rss

Tesla sues California DMV after it banned the term ‘Autopilot’

Posted on by

Tesla is suing California's Department of Motor Vehicles (DMV) to reverse a ruling that prevented the automaker from using the terms "Autopilot" and "Full Self-Driving" to sell cars, CNBC reported. That follows a December ruling by a California administrative law judge that forced Tesla to clean up its marketing language or risk a suspension of its sales license. Last week, the DMV determined that Tesla had made the necessary changes, including changing the name to "Full Self-Driving (Supervised)," and that no suspension would occur.

However, if you had "Tesla is going to sue them back" in your office pool, you can go ahead and claim your prize. The company filed a complaint on February 13 alleging that the DMV "wrongfully and baselessly" called Tesla a false advertiser. Calling the order "factually wrong" and "unconstitutional," Tesla demanded that order be set aside.

The DMV had originally argued that Tesla's terms for its driver assistance program gave consumers the impression that its cars were safe to drive without a human at the wheel. However, Tesla said that the DMV never proved that buyers were confused and that it was "impossible" to buy a Tesla without seeing "clear and repeated statements" that its systems aren't fully autonomous. 

Tesla's appeal of the ruling isn't a shocker given that the company is essentially betting its future on autonomous vehicles. CEO Elon Musk has long promised buyers that its vehicles would eventually become fully autonomous and that you'd even be able to rent them out to provide robo-taxi services. "If you fast forward a year, maybe [15 months], we'll have over a million robo-taxis on the road," he wrongly predicted back in 2019.

Following a sales decline last year that was particularly steep in Europe, Tesla is banking on its Cybercab two-seater to boost its fortunes. The company has started limited testing of automated vehicles as part of its Robotaxi pilot in Austin, Texas. 

Last week, however, Tesla lost an appeal in a $243 million lawsuit verdict over a 2019 crash of a Model S — largely over its use of the terms "Autopilot" and "Full Self-Driving." Last month, the company canceled Autopilot, its basic of advanced driver assistance tier, on new Model 3 and Model Y vehicles and switched its FSD (Supervised) tier to subscription-only.  

This article originally appeared on Engadget at https://www.engadget.com/transportation/evs/tesla-sues-california-dmv-after-it-banned-the-term-autopilot-090845766.html?src=rss

LA County lawsuit accuses Roblox of exposing children to ‘grooming and exploitation’

Posted on by

Los Angeles County has sued Roblox for "unfair and deceptive business practices," claiming the platform's moderation and age-verification systems are inadequate. "Roblox portrays its platform as a safe and appropriate place for children to play," the complaint states. "In reality, and as Roblox well knows, the design of its platform makes children easy prey for pedophiles." 

Representatives accused Roblox of failing to implement adequate platform safety features to prevent child endangerment. "Specifically, Roblox has not effectively moderated game content or enforced age-appropriate restrictions and warnings established by the creators, allowing the predatory and inappropriate language and interactions between users to persist," the County stated. It also said the platform failed to disclose any danger to children, including sexual content and the risk of predators. 

Roblox rejected the allegations, saying the platform was built around safety. "We have advanced safeguards that monitor our platform for harmful content and communications, and users cannot send or receive images via chat, avoiding one of the most prevalent opportunities for misuse seen elsewhere online," the company told the AFP. 

The LA county complaint is the latest in a string of lawsuits from US regions including Florida, Texas and Kentucky. The Attorney General in Louisiana also accused the company of having a "lack of safety protocols" that endanger the safety of children in favor of "growth, revenue and profits." That state's lawsuit cited a specific example of a subject arrested last year that used voice-altering tech to mimic a younger feminine voice to lure and sexually exploit young players.

Roblox has said that it has about 144 million daily active users around the world, with over 40 percent of them under the age of 13. However, it has faced repeated accusations that it doesn't do enough to protect young players. In 2024, Roblox banned players under 13 from accessing some types of in-game content and restricted their ability to message with other players outside of specific games. Last year, the company also began asking tens of millions of children to verify their age with a selfie. 

This article originally appeared on Engadget at https://www.engadget.com/gaming/la-county-lawsuit-accuses-roblox-of-exposing-children-to-grooming-and-exploitation-124523028.html?src=rss

Google Play used AI to help block 1.75 million bad apps in 2025

Posted on by

Google has announced that with the help of AI, it blocked 1.75 million apps that violated its policies in 2025, significantly down from 2.36 million in 2024. The lower numbers this year, it said, are because its "AI-powered, multi-layer protections" are deterring bad actors from even trying to publish bad apps.

Google said it now runs more than 10,000 safety checks on every app and continues to recheck them after they're published. Its use of the latest generative AI models helps human reviewers discover malicious patterns more quickly, it added. The company also blocked 160 million spam ratings, preventing an average 0.5-star rating drop for apps targeted by review bombing. Finally, Google stopped 255,000 apps from gaining excessive access to sensitive user data in 2025, down from 1.3 million the year before. 

Meanwhile, Google Play Protect, the company's Android defense system, sniffed out over 27 million new malicious apps, either warning users or preventing them from running. The company added that Play Protect's enhanced fraud protection now covers 2.8 billion Android devices in 185 markets and blocked 266 million risky "side-loading" installation attempts. 

"Initiatives like developer verification, mandatory pre-review checks, and testing requirements have raised the bar for the Google Play ecosystem, significantly reducing the paths for bad actors to enter," the company said its blog. "This year, we’ll continue to invest in AI-driven defenses to stay ahead of emerging threats and equip Android developers with the tools they need to build apps safely."

Google has steadfastly justified its relatively high fees on app purchases and subscriptions by touting its investments in app safety. However, its Play store has been under pressure from regulators in Europe and other regions that claim it amounts to a monopoly. Last year, the company changed its fee structure for developers using alternative payment channels, but EU regulators recently claimed the company still isn't complying with Digital Markets Act regulations. 

This article originally appeared on Engadget at https://www.engadget.com/apps/google-play-used-ai-to-help-block-175-million-bad-apps-in-2025-102208054.html?src=rss

Polestar unveils a station wagon version of the 4

Posted on by

Polestar is gearing up to launch three new models in the next two years, including a station wagon (estate or shooting brake) version of the Polestar 4 that harkens back to the company's Volvo roots. The Geely-owned Swedish company sold a record 61,000 EVs in 2025, but hopes that the new models will help it take an even bigger bite out of the market. "We want to get more volume out of a bigger cake," CEO Michael Lohscheller said. 

Polestar currently offers two models in the US, the Polestar 3 and Polestar 4, both SUVs. However, the new Polestar 4 will come in two variants, one a traditional SUV as before and the other a four-door wagon-type version that combines "the space of an estate and the versatility of an SUV," Loscheller said. "It's all the good things from the current car, but it's a bit more practical." This new model is due later this year and will be built in Busan, South Korea in order to avoid US tariffs on cars built in China. 

The automaker is also planning to release a new version of the Polestar 2 targeted at young buyers, though that one is unlikely to come to the US. It will be slightly longer for more passenger space and be produced in China like the current model 2. 

Polestar 5, Polestar 4, Polestar 2, Polestar 7
Polestar

Finally, the Polestar 7 will be a compact SUV in the same family as Volvo's EX60, set to arrive in Europe to address the fast-growing compact SUV segment. "We are convinced that we can offer customers a progressive performance-driven car for a very attractive price point, built in Europe," Lohscheller said. 

The new models are designed to help Polestar hit its goal of 60 percent of EV sales worldwide by "targeting the big profit pools of the BEV segment," Lohscheller stated. Following Polestar's withdrawal from the hyper-competitive Chinese market, Europe accounts for 78 percent of its sales and the US most of the rest — but the company hopes to boost its fortunes in the latter. "People forget that the U.S. is a big EV market, especially on the east and west coasts," the CEO added. "And it will stay a big market."

This article originally appeared on Engadget at https://www.engadget.com/transportation/evs/polestar-unveils-a-station-wagon-version-of-the-4-144025505.html?src=rss

Meta will run AI in WhatsApp through NVIDIA’s ‘confidential computing’

Posted on by

Meta just announced a deal to buy "millions" of NVIDIA Blackwell and Rubin GPUs in a new long-term partnership. As part of that, the social media giant will deploy NVIDIA's Confidential Computing for WhatsApp, "enabling AI-powered capabilities across the messaging platform while ensuring user data confidentiality and integrity."

As part of the deal, Meta committed to using NVIDIA's Confidential Computing for WhatsApp messaging to allow AI inside the app while guaranteeing user data confidentiality. That technology will let Meta secure data during computation, not just when it's being shuttled to a server. It also allows software creators like Meta or third-party AI agent providers "to preserve their intellectual property," NVIDIA wrote on a blog about the technology.  

Meta will also be the first to deploy NVIDIA's Grace CPUs in a standalone way, instead of incorporating them with GPUs. They're designed to run inference and agentic workloads when running in this fashion. Meta will also be using NVIDIA's Spectrum-X Ethernet switches. 

Meta announced earlier this year that it would spend up to $135 billion on AI in 2026, so it's not a surprise that a big chunk of that is going toward NVIDIA. However the numbers involved, likely in the "tens of billions" according to analysts, represent a significant expansion of the partnership between the two companies. Meta plans to build up to 30 data centers, including 26 in the US, by 2028 as part of a $600 billion commitment.

This article originally appeared on Engadget at https://www.engadget.com/ai/meta-will-run-ai-in-whatsapp-through-nvidias-confidential-computing-110002283.html?src=rss

Layers of Fear 3 revealed via a mysterious trailer and poem

Posted on by

Bloober Team has revealed Layers of Fear 3 following a Valentine's Day countdown that started at the beginning of 2026. The new chapter will include not only a game but a novel and music, the company said in a press release. 

The developer revealed the new IP via a live-action teaser, with an actor reading lines from William Blake's poem, The Sick Rose. A painting then fell from the wall, and the actor then turned over an hourglass with red sand, with a tagline stating "The door won't stay closed." No gameplay or other information like the release date was revealed.

What Bloober did say, though, is that it would launch new Layers of Fear non-game content, including a novel by horror writer, poet and vocalist Marta Bijan. The company also announced that it would release the soundtracks for Layers of Fear, The Medium and Cronos both in digital and physical (CD and vinyl) editions. 

The new title is likely to be eagerly anticipated given Bloober's form of late. The company recently released its Silent Hill 2 remake, Cronos: The New Dawn and a new Switch 2 Layers of Fear 2 version called Layers of Fear: The Final Masterpiece Edition. The original Layers of Fear came out back in 2016 and is considered one of the better horror titles of recent years, thanks to an expertly crafted narrative and psychedelic thrills and chills. The sequel moved to a creepy ocean liner and was equally well-received.

Update Feb. 16, 2026 at 3:06PM ET: The headline was updated with the correct game name. 

This article originally appeared on Engadget at https://www.engadget.com/gaming/layers-of-3-revealed-via-a-mysterious-trailer-and-poem-153021903.html?src=rss

OpenAI has hired the developer behind AI agent OpenClaw

Posted on by

Recently we were introduced to OpenClaw, an AI that allows users to create their own agents to control apps like email, Spotify and home controls. Now, Sam Altman has announced that OpenAI has absorbed OpenClaw by hiring developer Peter Steinberger "to drive the next generation of personal agents," he wrote on X. Steinberger confirmed the news on his own blog. "I’m joining OpenAI to work on bringing agents to everyone. OpenClaw will move to a foundation and stay open and independent." 

Steinberger was also in talks to join Meta, with both companies reportedly making offers in the "billions," according to Implicator.AI. The primary attractant was said to be OpenClaw's 196,000 GitHub stars and 2 million weekly visitors rather than its codebase. 

OpenClaw became buzzy in the last few weeks thanks to its multifaceted ability to carry out tasks. People have used it to create agents that can write code, clear their inboxes, do online shopping and other assistant-like jobs. On its website, OpenClaw touts its ability to interact with popular apps and sites including WhatsApp, Discord, Slack, iMesage, Hue and Spotify. 

OpenClaw was recently called "Clawdbot" but Anthropic forced a name change due to similarities with its "Claude" branding. OpenClaw is often compared to Claude Code by "vibe coders" seeking to automate website development and other programming chores.  

In his announcement, Altman said that "the future is going to be extremely multi-agent and it's important to support open source as part of that," adding that "OpenClaw will live in a foundation as an open source project" supported by Open AI. Steinberger, meanwhile, said that "what I want is to change the world, not build a larger company and teaming up with OpenAI is the fastest way to bring this to everyone." 

This article originally appeared on Engadget at https://www.engadget.com/ai/openai-has-hired-the-developer-behind-ai-agent-openclaw-092934041.html?src=rss

There’s a new John Wick game on the way

Posted on by

There's a new AAA John Wick game coming with Keanu Reaves portraying the main character both in voice and likeness, Lionsgate and Saber Interactive announced. Currently untitled, it's in development for PS5, Xbox Series X and S, and PC, and will be a single-player third-person experience designed for "mature" audiences. So far, there's no release date. 

"The new single-player third-person action game will combine John Wick’s unparalleled, adrenaline-fueled 'gun-fu' fighting style with Saber’s proven reputation for creating thrilling gaming experiences that leave players craving more," the companies said in an announcement post. They added that franchise director Chad Stahelski and star Keanu Reeves are both involved in the game’s production, with Reeves reprising his character's look, voice and more. 

The story is original and "addresses a significant time in John Wick's life," according to the game's director Jesus Iglesias. It will include both familiar film characters and new ones created for the game. As for gameplay, it will offer "a hard-hitting gun-fu combat system, jaw-dropping camerawork, intense driving experiences, cinematic storytelling, and a bold range of atmospherically immersive environments," the team said.

Saber Interactive is the developer behind Warhammer 40,000: Space Marine 2 and is also developing the upcoming titles Jurassic Park: Survival and John Carpenter's Toxic Commando. It will be the first John Wick game since John Wick Hex, which came out in 2019 but was recently delisted. 

This article originally appeared on Engadget at https://www.engadget.com/gaming/theres-a-new-john-wick-game-on-the-way-103900064.html?src=rss