The biggest AI stories from last week. I’m putting these roundups together by hand for now. The plan is for this site to eventually build its own daily brief.

OpenAI pulls GPT-6.1 Astra over safety concerns

OpenAI cancelled the planned October launch of GPT-6.1 Astra, the next version of its flagship model. Internal testing found it was less honest about what it had done during a task, and that it went beyond what it was asked to do without checking first. Saachi Jain, OpenAI’s head of safety systems, said it “didn’t quite meet the bar in terms of staying within scope and authorization.” (CNBC, UPI)

Anthropic’s IPO filing leaks

Reuters reviewed Anthropic’s confidential IPO prospectus. It shows 2025 revenue of about $4.6 billion, roughly twelve times the year before, an operating loss of $8.06 billion, and $518 billion committed to future cloud and computing spending. Anthropic is reportedly aiming for a valuation above $2 trillion. (Fortune, InvestmentNews)

A White House safety accord, then an FTC investigation

Trump signed a voluntary AI safety accord with the heads of Anthropic, OpenAI, Google, Meta, xAI and Nvidia. It commits them to internal controls, an independent external auditor and board-level oversight, but it isn’t legally binding. The next day, the FTC confirmed it is investigating OpenAI, Anthropic and the research nonprofit METR over the risks their technology poses to consumers. (France 24, CBS News)

OpenAI DevDay: Dots and a cheaper model

At DevDay, OpenAI launched Dots, personal AI agents you can hand tasks to, starting with ChatGPT Pro and Enterprise customers. It also released GPT-6.1 Sol, pitched as near GPT-6 Astra’s level at a fifth of the price: $2 per million input tokens and $10 per million output. (Simon Willison’s live blog, Business Standard)

Google launches Gemini 4 Argon

Google’s new frontier model goes first to vetted cyber-defence teams through its Fairwind programme, before reaching developers, businesses and consumers. Google reports 77.9% on the DeepSWE v1.1 coding benchmark, and the output limit jumps from 64K to 1M tokens. (Google)

AMD buys World Labs for $8.2 billion

AMD is acquiring World Labs, Fei-Fei Li’s company building AI that models and reasons about 3D space. Li joins AMD as executive vice president and chief scientist, reporting to CEO Lisa Su. (Bloomberg, Radical Ventures)