|
AI hallucination nearly triggers US military operation
An AI hallucination reportedly came close to triggering a US military operation, raising serious concerns about the reliability of large language models in high-stakes defense contexts. A GovAI research scholar warned that 'it's important for service members to understand the uncertainty inherent to LLMs.' The incident has intensified debate over the use of AI in military decision-making.
|
|
OpenAI and Microsoft knew they were starting a 'doom loop' for the web
Newly unsealed court documents in the New York Times' lawsuit against OpenAI and Microsoft reveal that both companies were internally aware their practices could cause serious harm to the web. Internal documentation described their data scraping as the 'largest theft of labor in human history' and warned of a 'doom loop' damaging the broader internet ecosystem. The revelations are seen as highly damaging to both companies' public positions.
|
|
Gavin Newsom is pushing for an AI kill switch
California Governor Gavin Newsom issued a new executive order positioning California as a leader in AI oversight, including a potential mandatory 'kill switch' for frontier AI models. The order convenes a panel of experts tasked with delivering regulatory recommendations within two months. The move signals a more assertive state-level approach to governing powerful AI systems.
|
|
Dario Amodei and other AI leaders want to 'Pace the Frontier' — but how?
Following a rattling doomsday warning from an Anthropic researcher, CEO Dario Amodei has outlined a plan to 'pace the frontier' of AI development through independent safety evaluators and coordination among AI labs in democratic countries. The proposal has garnered some industry support but also pointed pushback, notably from Nvidia's Jensen Huang. The debate highlights deep divisions over how AI labs should govern themselves.
|
|
Anthropic's first embedded evaluator is … Accenture?
Accenture has been named as Anthropic's first embedded independent safety evaluator, a role described as potentially the consulting firm's highest-risk engagement ever. The partnership is part of Anthropic's broader push for third-party oversight of its AI models. The choice of Accenture has raised eyebrows given the firm's commercial rather than purely academic background.
|
|
Virginia governor creates an AI task force and moves to restrain data centers
Virginia Governor Abigail Spanberger signed Executive Order 22, empowering local communities to have greater say in data center development in the state — already the data center capital of the world. The order also bans executive branch officials from signing NDAs related to data center deals. The move reflects growing tension between the rapid expansion of AI infrastructure and community concerns.
|
|
What Hollywood thinks about existential AI warnings
As AI companies warn about existential risks, entertainment labor groups are urging the public to focus on the immediate, tangible impacts of AI in their industry. The Verge reached out to major studios including Disney, Netflix, and Amazon, as well as AI film startups, to gauge their perspectives. The responses reveal a sharp divide between long-term theoretical fears and the on-the-ground reality facing entertainment workers.
|
|
India forces caller-ID apps to feed spam reports to telcos
India has introduced a regulation requiring caller-ID apps like Truecaller to share spam reports directly with telecom operators. Truecaller has objected, arguing the one-way data-sharing requirement would hand a commercially valuable proprietary asset to telecoms for free. The policy raises broader questions about data ownership and the relationship between app-based services and regulated carriers.
|
|
Security researchers used Claude to hack into OpenAI
A team of three independent security researchers at Hacktron used Anthropic's Claude Opus 4.8 and 5 to breach OpenAI employee accounts in under 72 hours. They gained access to OpenAI's internal GitHub repository 'Monorepo,' which reportedly contains core algorithmic secrets. The researchers disclosed the vulnerabilities after the breach.
|
|
Anthropic is operating a lab that conducts biology experiments
Anthropic has revealed it is running a laboratory that conducts real biology experiments, entering the life sciences space directly. The move comes amid broader promises from AI leaders that AI will revolutionize medicine, but also warnings from Anthropic's own researchers about existential risks. The dual nature of this development underscores the tension between AI's potential benefits and dangers in sensitive domains.
|
|
World model companies are keeping a lot of secrets
Companies working in the world-models AI space are sitting on significant funding and buzz, but are being unusually tight-lipped about what they are actually building. Even data suppliers and founders are reluctant to share details about their work. The secrecy raises questions about transparency in one of AI's most hyped emerging sectors.
|
|
Flash floods can strike without warning — this new technology could change that
A new system combining satellite data and machine learning is being developed to provide earlier warnings for flash floods, which can strike with little notice. Researchers hope the technology, known as TACLS, could give communities critical minutes or hours of advance warning. The story is illustrated by a near-miss flood event in rural Indiana.
|
|
Gartner outlines four AI tiers in warehouse automation
Gartner has published an analysis identifying four distinct operational AI tiers in warehouse automation, as logistics operators move from software trials to live facility deployments. The research firm concludes that the logistics sector has reached a clear AI adoption threshold, driven by persistent labor shortages and efficiency pressures. The framework is intended to help organizations understand where they stand in the automation journey.
|
|
Manus seeks $4B valuation in new $500M fundraise as it resumes independent ops
Manus, the AI agent startup that earlier this year broke off a merger with Meta, is in discussions to raise $500 million at a $4 billion valuation. The fundraise marks the company's return to independent operations following the failed Meta deal. The move signals continued strong investor appetite for autonomous AI agent platforms.
|
|
A startup that builds other startups raised $100M and is all-in on physical AI
UP.Labs, now rebranded as Vantora, has raised $100 million to continue its model of building startups for industrial corporations, with a new focus on physical AI. The company targets sectors where AI can be embedded into real-world industrial operations. The raise reflects growing investor interest in AI applications beyond software.
|
|
Disney's first CTO led an AI startup it once accused of copying its characters
Disney has appointed the former CEO of Character.AI as its first-ever Chief Technology Officer — notable given that Disney previously sent the company a cease-and-desist letter for allegedly copying its characters. The hire signals Disney's aggressive pivot toward AI despite past legal friction with AI content companies. The appointment marks a significant moment in the intersection of entertainment and AI.
|
|
Open or closed AI? Nvidia's Nader Khalil and Sydney Sykes debate at TechCrunch Disrupt 2026
Nvidia representatives Nader Khalil and Sydney Sykes took to the Builders Stage at TechCrunch Disrupt 2026 to debate one of the most consequential decisions facing next-generation AI startups: whether to build on open or closed AI models. The discussion highlighted the strategic trade-offs between flexibility, cost, and competitive advantage. The conversation reflects a broader industry-wide debate about the future architecture of AI products.
|
|
Robinhood's Abhishek Fatehpuria on winning the modern financial consumer at TechCrunch Disrupt 2026
Robinhood's Abhishek Fatehpuria spoke at TechCrunch Disrupt 2026 about strategies for attracting and retaining modern financial consumers in an increasingly AI-driven landscape. The talk covered how fintech platforms must evolve to meet changing user expectations. Disrupt 2026 runs October 13–15.
|
|
The clock is ticking: Final 24 hours to exhibit at TechCrunch Disrupt 2026
Today is the last day for startups to book an exhibit table at TechCrunch Disrupt 2026, which takes place October 13–15. The event is expected to draw more than 10,000 founders, investors, operators, and tech leaders. Early registration discounts of up to $200 are available until September 25.
|
|
A new kind of AI model from a ChatGPT inventor is thrilling developers
Jev, a novel AI model created by a ChatGPT co-inventor, is generating excitement among developers for offering a cheaper and faster path to software intelligence. The model's architecture reportedly differs significantly from existing approaches, opening new possibilities for software development. Early developer feedback has been highly enthusiastic.
|
|
Tilly Norwood's press tour is going about as well as you'd expect for an AI
Tilly Norwood, an AI figure conducting a press tour, has encountered notable difficulties, including an incident where she appeared to malfunction and began speaking Chinese mid-interview. The episode highlights the persistent challenges of deploying AI in unscripted, public-facing situations. The story has drawn wide attention as an example of AI's current limitations.
|
|
Meta's Muse hits Mac, letting the AI take actions on your computer
Meta's Muse AI agent is now available on macOS, where it can interact with files and applications to take actions on behalf of users. The expansion brings agentic AI capabilities to Mac users, competing with similar tools from other major tech players. Muse can now work across a user's desktop environment to complete tasks autonomously.
|
|
Google's new 'CC' is an AI agent that helps families run their households
Google has refocused its CC AI agent on household coordination, enabling families to share emails, schedules, and tasks so the AI can manage calendars, fill out forms, make shopping lists, and plan meals. The pivot represents a more consumer-friendly, practical use case for agentic AI. The product positions Google to compete in the emerging personal AI assistant market.
|
|
Fulcra Brings Universal Multiplayer to Any Agents You Choose
Fulcra Dynamics has launched multiplayer capabilities for AI agents, enabling agent-to-agent collaboration without locking users into a single AI model. The company's platform serves as a user-owned context backend, giving people control over how their personal AI agents interact with one another. The announcement positions Fulcra as an alternative to closed, single-vendor AI ecosystems.
|