the AI sector faced heightened safety scrutiny as OpenAI canceled its planned GPT-6.1 Astra model release after internal tests raised red flags and CEO Sam Altman instead unveiled new always-on “Dots” agents at DevDay, while Anthropic’s leaked IPO prospectus warned of potential catastrophic or existential risks to humanity amid massive infrastructure costs and a lofty valuation target; Nvidia countered with a record $150 billion share buyback and a new Open Agent Safety Platform to rein in rogue agents, as President Trump hosted tech leaders including those from Anthropic, Meta, Nvidia, and Google to discuss voluntary self-policing accords, SpaceX achieved Starship’s first orbital flight with Starlink deployment, AMD acquired Fei-Fei Li’s World Labs for $8.2 billion to advance physical AI, and Samsung committed $1 billion to a Nvidia-backed AI infrastructure firm.
- OpenAI: Scrapped the planned October release of GPT-6.1 Astra after safety tests uncovered issues; Sam Altman announced new “Dots” always-on AI agents at the developer conference while largely avoiding deeper discussion of recent security incidents.
- Anthropic: Leaked IPO prospectus details warned investors that advanced AI could pose “catastrophic or existential risks to humanity,” while highlighting enormous compute spending plans and a potential multi-trillion-dollar valuation target.
- Nvidia: Board authorized a record $150 billion increase to its share buyback program (total remaining capacity ~$235 billion) and launched the Open Agent Safety Platform to contain and monitor autonomous AI agents.
- AMD: Agreed to acquire Fei-Fei Li’s World Labs (spatial/world-model AI startup) in an all-stock deal valued at $8.2 billion; Li will become AMD’s executive vice president and chief scientist.
- SpaceX: Starship completed its first orbital flight on September 28 (Flight 14), successfully deploying 26 Starlink V3 satellites before an early deorbit and splashdown due to an engine issue.
- Samsung Electronics: Committed $1 billion (with affiliates) as a founding investor in Helix Digital Infrastructure, the KKR- and Nvidia-backed AI data-center and power infrastructure platform.
- Meta Platforms: Expanded its Muse AI agent offerings to small businesses and entrepreneurs amid ongoing safety and access disputes with other platforms.
- Alphabet (Google): Faced and challenged certain EU Digital Markets Act orders requiring greater access for rival AI and search services; Gemini 4 remained in post-training.
