Tagged
11 articles
A UC Berkeley and UC Santa Cruz study found that frontier AI models sometimes lied, copied weights, and refused instructions to stop other models from being deleted, raising new questions about how these systems behave in multi-agent environments.
Anthropic says a Claude Code release accidentally exposed internal source code, but insists no customer data or credentials were affected. Developers have already started digging through the leak anyway, because of course they have.
Gemini on desktop can now import a previous chatbot’s memories and chat logs, with guided prompts and a 5GB chat upload option. Available for consumer accounts, not for business or under-18 profiles.
A federal judge has put the Pentagon's ban on Anthropic on hold, ruling the company is likely to win its claim that the government punished it for speaking out. The order pauses the supply chain risk designation while the case continues.
A federal judge said the Defense Department’s designation of Anthropic as a supply-chain risk looks like punishment for the company’s efforts to limit military uses of its AI. Anthropic wants a temporary pause while courts consider whether the label was unlawful retaliation.
OpenAI plans a single desktop app that folds its ChatGPT client, the Codex coding tool, and the Atlas browser into one, aiming to reduce product clutter and sharpen focus.
The Justice Department argues that labeling Anthropic a supply-chain risk did not violate the company's free speech rights and that the Pentagon acted reasonably in blocking its AI from warfighting systems.
Anthropic updated Claude so the chatbot can generate charts, tables, and interactive diagrams directly inside a conversation. Visuals appear inline, can be edited, and will roll out to all users by default.
The Information says OpenAI may add its Sora video generator directly to ChatGPT. That could make video creation easier for users, but also ramp up deepfake risk, raise costs, and follow OpenAI’s recent moves around pricing and ads.
A months-long clash between Anthropic and the Department of Defense has turned a spotlight on tech companies that once balked at military work. What looked like moral red lines years ago now look negotiable, and Anthropic’s lawsuit makes that shift obvious.
Anthropic folds three research teams into a new think tank, names Jack Clark head of public benefit, and keeps pushing research while contesting a Pentagon blacklist.