What’s changing
in AI and work.

The FCC weighs AI-generated political robocalls, OpenAI pushes agents into team coordination, AMD plans a 2027 supply increase, and a Delaware suit challenges Nvidia and Groq’s licensing-and-hiring deal.

Mistral previews its ML4 open-weight model, leading AI labs stop short of absolute agent-safety guarantees, Taiwan indicts 10 people in an alleged military-chip export scheme, and workplace research finds changed tasks without measurable earnings or hours effects.

A federal investigation puts AI-agent safety under scrutiny, courts reshape the rules for AI Overviews and synthetic images, campaigns spend millions on AI-generated ads, and Reflection AI unveils its open-weight Beam model.

An OpenAI safety leader resigns with a warning, Trump announces a “Super Intelligence Force,” a McKinsey report estimates 11 million U.S. workers may need to change occupations by 2035, and deepfake detection gets harder.

Brazil set election rules for AI-generated political content, President Trump backed a voluntary six-company safety pact without penalties, Apple promised tighter Mac permissions for AI agents, and a China-aligned group reportedly used AI-policy impersonation to steal credentials.

A Guardian report says a $20 AI subscription helped expose roughly 1.52 million Georgia voters’ ballots, while AI-aligned super PAC spending reached $55.7 million. Google recapped Gemini 4 Argon, Releasebot reported a wave of ChatGPT updates, and FieldAI pursued a $700 million round at a $10 billion valuation.

OpenAI’s DevDay brought always-on Dots agents, GPT-6.1 Sol and Codex Cloud; Google unveiled Gemini 4 Argon; court and regulatory actions raised the stakes for AI training and autonomous agents; and the UK government said it is preparing for possible labor-market disruption.