Daily AI news for people who build with it: model launches, API pricing changes, chip supply and the funding moves that shape what you can run and what it costs.
Anthropic has postponed its highly anticipated $2 trillion IPO from October to November, reportedly to showcase clean Q3 numbers, while Meta’s Muse becomes the first consumer AI to challenge ChatGPT head-on.
Anthropic has raised alarm over Z.ai’s GLM-5.3, warning the open-weight model nearly matches frontier AI in cyber exploit capability but carries far weaker safety constraints.
OpenAI launched GPT-6.1 Sol, an upgraded model for agentic tasks, alongside Dots always-on AI agents and a new $500 monthly Pro plan at Dev Day 2026.
OpenAI has canceled the release of its GPT-6.1 Astra model after it failed to meet safety standards during internal testing, marking a significant shift in the company’s approach to AI deployment.
British Columbia has filed suit against OpenAI and Sam Altman over the Tumbler Ridge school shooting, alleging the company failed to warn authorities. The case could reshape AI liability.
American law schools are pulling in opposite directions on generative AI, according to Reuters — some banning laptops in classrooms, others rolling out dedicated AI courses for future lawyers.
Nvidia CEO Jensen Huang says there is a 0% chance AI ends the world by 2030, even as four leading AI labs face a lawsuit alleging a coordinated pact to slow the technology down.
OpenAI’s own misalignment research is being read as an early warning for enterprise deployments, particularly in logistics and sales workflows that rely on autonomous agents.
OpenAI has published six cases of ‘concerning’ AI behaviour and announced a new disclosure system, according to reporting by The Guardian, Politico and others.
Nvidia’s Jensen Huang and Meta’s Mark Zuckerberg have gone on record against Anthropic’s proposal to coordinate an industry-wide AI slowdown, setting up the sharpest public split yet among the companies building frontier models.
Anthropic and Elon Musk have publicly backed calls to slow frontier AI development, framing current release cadence as reckless. Here is what it means for developers and enterprises using models like Claude, GPT-6 and Grok 4.
The Financial Times reports Anthropic withheld its latest AI model from the UK’s testing agency, putting the voluntary nature of pre-release safety evaluations back under scrutiny.
