Tuesday, Sep 15, 2026
AI;DR
2026-09-15- openaiOpenAI backs the bipartisan FRONTIER Act requiring outside AI safety evaluators, matching Anthropic’s pledge of permanent independent access; CEO Sam Altman supports federal safety rules but says firms need not wait for legislation.
- metaMeta committed to over 1 GW of custom AI silicon capacity in 12 months, deploying the MTIA 450 “Arke” chip in H1 2027 and MTIA 500 “Astrid” by year-end, built with Broadcom and TSMC to beat Nvidia on perf/watt.
- gates-foundationThe Gates Foundation will spend at least $1 billion over two years to expand AI access in underserved regions, with Bill Gates warning that unequal adoption could widen the global rich-poor divide.
- bytedanceByteDance’s H1 revenue hit $120B (+30% YoY), surpassing Meta’s $117B, with overseas TikTok operations now over 30% of revenue; net profit dipped on heavy AI spend across Doubao, Seedance and an in-house inference chip.
- anthropicAnthropic CEO Dario Amodei’s call to slow AI development for safety, backed by Sam Altman and Elon Musk, sparked a market split: cybersecurity stocks jumped 10–15% while chipmakers like Nvidia and Oracle fell ~5% on GPU demand fears.
- deepseekDeepSeek engineer Liu Shengyu clarified his viral essay mourned hand-written GPU kernels, not job loss, saying AI already outpaces him and he now leans on agents; he reaffirmed optimism about his role.
- nvidiaPresident Trump called AI takeover fears a “hoax” during a live on-stage call with Nvidia’s Jensen Huang, who dismissed AI doom predictions as unscientific; both framed AI as vital to US growth and security.
- anthropicFormer Anthropic researcher Jacob Coxon warned autonomous self-improvement could let AI outpace human control within 3–5 years, potentially threatening humanity by decade’s end.
92 stories