Gemini 4 Argon starts with cyber defenders
Google has announced Gemini 4 Argon, a model for long, multi-step work across software engineering, legal and financial tasks, and cybersecurity defense.
Access is narrow for now. Argon is rolling out to trusted cyber defenders through Google's Fairwind Program while the company tests safeguards. Google says a broader release will start with paid API customers and Google AI Ultra subscribers.
AI leaders agree to self-police
Executives from Google, Anthropic, Meta, OpenAI, xAI and Nvidia have signed an AI safety agreement. It calls for internal controls, dedicated safety teams, independent assessments and board-level oversight of frontier models.
The agreement says those measures could eventually become laws or regulations, but it specifies no penalties for violations today.
Reddit closes two doors
Reddit will end RSS support on November 13 and close its public API in March 2027, citing scraping and automated abuse.
The API change will affect research tools, social-listening products and AI assistants that access Reddit conversations programmatically. Approved third-party apps and bots must register by January 12, 2027 to avoid disruption.
Also
- Gemini gains reusable skills Save instructions once and invoke them with a slash command; the global Gemini chat rollout has begun, with Workspace availability coming in the next few weeks.
- OpenAI hack fallout: a lawsuit The nonprofit LASST seeks court restrictions on unauthorized system access and unsafe AI development practices. OpenAI calls the case meritless.
- OpenAI hack fallout: training monitors Research chief Mark Chen says all training runs are now monitored and 5% to 10% of computing resources have shifted from model training to safety work.
Pass it on