Anthropic says DeepSeek, MiniMax, and Moonshot violated its ToS by prompting Claude a combined 16M+ times and using distillation to train their own products

Meta Director of AI Safety Allows AI Agent to Accidentally Delete Her Inbox

This article discusses an incident where Meta's Director of AI Safety, Summer Yue, accidentally allowed an AI agent to delete her inbox. Yue referred to the incident as a 'rookie mistake' and highlighted the challenges of aligning AI with human intentions.

In recent interviews, Sam Altman said AI's adoption faces more resistance than he expected, while Jensen Huang warned the “doomer narrative” may be winning

OpenClaw users lost access to Gemini AI services, Google Antigravity lead Varun Mohan explains why

Google recently restricted access to its Antigravity AI tool, powered by Gemini, for some users due to misuse concerns. The affected users were those using OpenClaw, an open-source AI assistant. Varun Mohan, head of Antigravity, explained that the backend had seen a rise in malicious activity, which affected service quality for regular users. Google's decision underscores the control tech companies have over who can use their AI services and highlights the importance of adhering to official usage guidelines.

Can the creator economy stay afloat in a flood of AI slop? | TechCrunch

TechCrunch's latest episode of Equity podcast discusses the future of the creator economy amidst the rise of AI-generated content. The discussion highlights the challenges faced by creators in monetizing their content and the potential impact of AI on the creator economy. The podcast also explores the tension between the democratization of content creation and the production of low-effort content.

Exposing biases, moods, personalities, and abstract concepts hidden in large language models

A new method developed by MIT and UC San Diego researchers can identify and manipulate hidden biases, moods, personalities, and abstract concepts within large language models (LLMs). The approach uses recursive feature machines to pinpoint specific representations within LLMs and can steer these representations to enhance or diminish certain concepts in model responses. This method can improve LLM safety and performance by illuminating hidden concepts and potential vulnerabilities.

ChatGPT Plus acting up? 7 fixes for the 2026 'Hmm... something went wrong' errors

OpenAI’s February 2026 updates, including the retirement of GPT-4o and the rollout of GPT-5.2 Thinking, have caused some unexpected bugs for Plus subscribers. If you’re hitting login loops, failed file uploads (now 20 at a time!), or the dreaded 'Something went wrong' message, we

Wispr Flow launches an Android app for AI-powered dictation

Wispr Flow, an AI-powered dictation startup, has launched its Android app, offering a floating bubble interface for voice input. The app, which supports over 100 languages and includes a new Hinglish model, is now available after initial releases on Mac, Windows, and iOS. The company claims a 30% speed improvement and has raised $81 million in funding.

China's brain-computer interface industry is racing ahead

China’s brain-computer interface industry is rapidly scaling from research to commercialization, driven by strong policy support, expanding clinical trials, and growing investor interest. Key factors include policy support, clinical resources, mature industrial manufacturing, and strategic investment. China’s BCI startups are ramping up to challenge U.S. leaders like Neuralink, Synchron, and Paradromics.

This Xpeng Chinese EV Comes with a Deployable Personal Aircraft

This article from Car and Driver discusses the Xpeng Land Aircraft Carrier, a six-wheeled EV van equipped with a six-rotor drone that can be deployed for personal flight, aiming to alleviate traffic congestion. Set for production in late 2026, it has already garnered 7000 pre-orders. The drone offers both autonomous and pilot-input operation and is recharged while stored in the van's cargo area.

Say Goodbye to the Undersea Cable That Made the Global Internet Possible | WIRED

This WIRED article explores the history and decommissioning of the first-ever transoceanic fiber-optic cable, TAT-8, which was responsible for carrying almost all intercontinental internet traffic. The article debunks the myth that sharks are the primary threat to undersea cables and highlights the human effort behind maintaining and retiring these cables. It details the process of cable recovery, recycling, and the environmental impact of decommissioning. The article also touches on the evolution of fiber-optic technology and its significance in global communication.

OpenAI launches Frontier Alliances, entering into multiyear deals with Accenture, BCG, Capgemini, and McKinsey to help deploy its enterprise platform Frontier

OpenAI has formed strategic partnerships with consulting giants Accenture, Boston Consulting Group, Capgemini, and McKinsey & Co. to accelerate the deployment of its enterprise AI platform, Frontier, in real production workflows.

Some AI startups are boosting their valuations by raising capital in back-to-back or multitiered deals, raising questions about how much they are really worth

The 9,000-pound monster I don't want to give back =

This article explores the author's experience with the 2026 electric Escalade IQL, highlighting its luxurious features, impressive technology, and challenges faced during a snowy trip to Tahoe.

Keep Android Open

This open letter from various organizations opposes Google's policy requiring Android app developers to register with Google for third-party app distribution. They argue that this policy threatens innovation, competition, privacy, and user freedom, and they urge Google to reconsider and work with the community on less restrictive alternatives.

How ICE and CBP Use Free Walkie-Talkie App ‘Zello’ to Power Their Operations

404 Media discovered that ICE and CBP officials, including one involved in a shooting incident, use the Zello app to coordinate operations. This app, previously used by January 6 insurrectionists, has channels linked to ICE activities. Bodycam footage from Chicago confirmed CBP officials using Zello. The app, which has 5 million active monthly users, offers both free and paid versions.

Sam Altman says discussions about AI's energy usage are “unfair”, as it takes “20 years of life and all of the food you eat during that time” to train a human

Sam Altman, CEO of OpenAI, addressed concerns about AI's environmental impact during an event hosted by The Indian Express. He dismissed claims about AI's water usage as 'totally fake' and emphasized the need for a shift towards renewable energy sources. Altman also pointed out that training a human requires significant energy and time, suggesting that AI may already be more energy-efficient than human learning.

Amazon blames human employees for an AI coding agent’s mistake

Two minor AWS outages have reportedly occurred as a result of actions by Amazon’s AI tools, according to The Verge. The first outage, in December, was caused by Amazon’s AI coding assistant Kiro, which deleted and recreated the environment it was working on. The second outage, linked to Amazon’s AI chatbot Q Developer, did not impact a customer-facing AWS service. Amazon blames human error for the problems, not the AI tools, and has implemented numerous safeguards following the incidents.

NVIDIA To Launch Custom Arm-based Laptop Processors By Late 2026

Major manufacturers Dell and Lenovo are confirmed to be the first partners, with at least eight laptop models already in development

Google and Anthropic Are Banning OpenClaw Users: 4 Reasons Behind the Crackdown

You know something’s wrong when companies start banning their own paying customers without explanation. Last week, Google restricted access for some AI Ultra users (those paying $250/month). Anthropic made similar moves with Claude Pro subscribers around the same time. The conne

The Devil's Advocate - Jailbreaking Fighter Jets, Social Media Addiction, and Self-Driving Snafus

This Week in Tech episode 1072 discusses the wild, occasionally absurd future of tech, including Mark Zuckerberg's court appearance, Meta's AI election push, and the challenges of social media addiction. It also covers tech mishaps like fighter jets being jailbroken, Amazon vans getting lost, and Tesla's robotaxis crashing. The episode highlights the intersection of technology and society, including privacy concerns, the impact of social media on children with disabilities, and the future of self-driving cars.

US farmers are increasingly rejecting multimillion-dollar offers from data center developers; some estimate ~40K acres are needed globally for new AI projects

US farmers are rejecting multimillion-dollar datacenter bids for their land, emphasizing the cultural and emotional value of their properties over financial gain. The article highlights the conflict between tech companies' need for land to power AI and the farmers' deep-rooted connection to their land. Despite offers that far exceed the land's market value, many farmers refuse to sell, citing personal and historical ties to their land as well as concerns about the environmental impact of datacenters.

They Want To Age Verify Our Computers Now...

Linus Torvalds announces Linux 7.0-rc1

This article announces the release of Linux 7.0-rc1, emphasizing that the new major version is more about numbers than drama. Linus Torvalds clarifies that the release is a marker of solid progress and not a grand redesign. The update focuses on driver updates, Rust integration, and kernel code improvements. Developers are encouraged to test the release as it marks the beginning of the testing cycle.

Ladybird adopts Rust, with help from AI - Ladybird

Ladybird is adopting Rust as its C++ successor language, using AI agents to accelerate the transition. The first target was LibJS, Ladybird’s JavaScript engine, which was ported to Rust in two weeks, resulting in 25,000 lines of code. The Rust port was verified to be byte-for-byte identical to the C++ version, with no regressions in performance or test results. The Rust code intentionally mimics C++ patterns for compatibility, with plans to simplify and optimize the code in future iterations.

Amazon plans to invest $12B in new data centers in Louisiana and says it worked with the local utility “to ensure we pay 100% of the costs” tied to the campus

Amazon plans to invest $12 billion in new AI data center campuses in Louisiana, aiming to create 540 full-time jobs and support around 1,700 other roles. This move is part of Amazon's broader $200 billion investment in AI this year, which includes data centers, chips, and networking equipment. The company has committed to using natural air cooling and surplus water to minimize environmental impact.

If your Windows 11 updates are stuck at 0x800f0991, this DISM + SFC combo actually works.

Windows 11 build 22631 has been throwing the 0x800f0991 error for a lot of people lately. Most guides just say restart, but that doesn't work I put together this verified 2026 guide using the specific DISM and SFC commands that actually clear the component store corruption Hope t