⚖️ NYT reporter John Carreyrou sues major AI companies over book training
John Carreyrou and several other writers filed suit alleging Anthropic, Google, OpenAI, Meta, xAI, and Perplexity trained models on pirated copies of their books. The sharper tactical note is their decision to avoid a class action, arguing that class treatment can compress many claims into one cheaper settlement path.
🧠📜 New York and California start regulating “AI companions”
New York’s law is live and leans on repeated “I’m not human” disclosures, plus detection and referral when users express suicidal ideation. California’s framework goes broader, with heavier youth protections and stronger requirements around prevention, transparency, and reporting.
🏥🔍 Rolling back health AI transparency rules could dump work on hospitals
A STAT piece flags a proposal that would loosen disclosure expectations for clinical AI tools, shifting the diligence burden onto health systems. The practical consequence is more vendor-by-vendor excavation of validation details, and fewer standard “receipts” when performance questions surface.
🏛️🤖 UK government publishes results on an AI tool that reads consultation responses
The UK Department for Transport released an evaluation of its Consultation Analysis Tool (CAT), built with the Alan Turing Institute, to do thematic analysis across large volumes of free-text consultation responses. The story here is administrative scale: human review remains central, while the tool accelerates clustering and theme discovery.
🏠🗣️ Amazon’s Alexa+ adds more service integrations
Amazon is extending Alexa+ into “assistant as service gateway” territory with integrations such as Angi, Expedia, Square, and Yelp. The bet is that conversational back-and-forth becomes the interface for booking, quotes, and scheduling, rather than app switching.