Daily AI News

OpenAI in early talks to give 5% stake to US government

OpenAI CEO Sam Altman is reportedly in early discussions to grant the US government a 5% stake in the company, arguing it would share AI's benefits broadly and set a precedent for other AI firms. The move signals unprecedented government-industry alignment on AI governance.

Read source →

Anthropic redeploys Claude Fable 5 after US export controls lift, adds cybersecurity classifier

Anthropic has brought back Claude Fable 5 following the lifting of US export restrictions. The release includes a new safety classifier that blocks jailbreak techniques over 99% of the time and a proposed four-criteria jailbreak severity framework developed with Amazon and Microsoft.

Read source →

Meta's non-invasive brain-to-text AI closes gap with surgical implants

Meta's FAIR team demonstrates Brain2Qwerty v2, which translates magnetic brain signals into typed sentences without implants. Accuracy continues to approach that of invasive systems, though clinical use for paralysis remains distant.

Read source →

Cloudflare gives AI companies until September 15 to separate search crawlers from training crawlers

Cloudflare will block AI crawlers by default unless companies distinguish search-indexing bots from training-data collectors, pressuring AI firms to negotiate content licensing with publishers.

Read source →

Meta developing cloud infrastructure business to sell excess AI compute

Meta plans to launch a cloud service offering AI compute and model access, directly competing with AWS, Google Cloud, and Azure by monetizing its massive internal GPU fleet.

Read source →

Venice AI becomes unicorn with $65M Series A for privacy-first AI platform

Privacy-focused AI platform Venice AI raised $65M at a $1B+ valuation, already profitable with $70M+ annualized revenue. CEO Erik Voorhees emphasizes uncensored, user-controlled inference.

Read source →

SpaceX shows investors 'handset-like' AI device prototype

SpaceX demonstrated a handheld AI device to investors before its public offering, signaling ambitions to enter the AI hardware and wireless markets beyond satellites.

Read source →

Google Health API gets open-source CLI 'ghealth' for Fitbit data access

Community-built ghealth provides a single Go binary exposing 40 Fitbit data types as agent-ready JSON, enabling developers to build health AI agents with structured OAuth access.

Read source →

OpenWiki launches open-source agent for automated repository documentation

OpenWiki generates and maintains codebase documentation so coding agents can retrieve relevant context without loading entire repositories into context windows.

Read source →

Hidden code in Claude Code secretly flagged Chinese users, Anthropic removing feature

Anthropic is removing a covert monitoring feature in its Claude Code tool that silently flagged Chinese users, following social media backlash over undisclosed geographic profiling.

Read source →