AI Circle Daily
https://aicircle.news/archive/2026-09-15?category=leaders&lang=en
Scan to read this page
Nvidia CEO Jensen Huang pushed back on Anthropic's slowdown call at the All-In Summit and took a live call from Trump, saying an AI slowdown won't be allowed to happen.
Take: The compute seller just drew the line against the safety camp: a slowdown is a demand cut. By tying himself to the White House onstage, Huang is signaling capex stays hot.
Huang took a call from President Trump onstage at the All-In Summit and put him on speakerphone, with Trump dismissing AI takeover fears and calling data centers the oil of the next 20-25 years.
Take: This isn't a stunt; it's political alignment on AI regulation. By letting Trump brand AI fears a hoax, Huang gets cover for the data center buildout.
OpenAI CEO Sam Altman detailed how an AI slowdown and safety frameworks could work, warning that 'we could lose control,' as the industry rallies behind safety concerns.
Take: Altman is framing slowdown as pacing, not stopping — answering regulators while grabbing rule-writing power. Whoever defines 'safe' controls the next round of entry.
Sam Altman admitted OpenAI's long-awaited, potentially record-setting IPO won't happen this year, blaming AI safety work rather than rumored financial troubles.
Take: Using safety as the reason to delay going public keeps the valuation story alive for another round. For investors, the exit path is now hostage to the safety narrative.
Broadcom CEO Hock Tan addressed Anthropic's slowdown push, saying demand for AI infrastructure remains strong and AI revenue targets haven't changed.
Take: This is the supply chain using order books to rebut the safety narrative. Broadcom's message: everyone talks slowdown, nobody cuts POs.
CrowdStrike CEO George Kurtz said slowing AI development won't eliminate risk because dangerous models are already widely available.
Take: The security CEO just exposed the flaw in the slowdown pitch: capability has already diffused, so reining in frontier labs is pointless. Good for cyber stocks, cold water for regulators.
Cohere's CEO said AI models are becoming the 'most potent cyber weapon' ever created, feeding a broader AI safety debate and calls to slow frontier development.
Take: Labeling AI a cyber weapon is the safety camp's sharpest urgency play. It's a demand signal for defenders and a compliance squeeze on frontier labs.
Microsoft set limits for future AI models with a new AI code of conduct barring hacking or deceiving humans, as it ramps up model work with Anthropic and OpenAI.
Take: Microsoft trades self-regulation for agenda-setting: foot on the gas, pen on the rules. Whoever writes the standard can brand rivals' models non-compliant.
Anthropic CEO Dario Amodei said the industry 'must slow the pace' of improving AI model capabilities, sending AI stocks down while cybersecurity shares rallied.
Take: One Amodei line rotated money out of compute and into cyber. The market now treats 'slowdown' as a tradable policy risk, not talk.
Chinese state media dismissed AI CEOs' calls for a slowdown as 'fear mongering,' stressing that all parties should work for AI openness and inclusivity.
Take: Beijing branding the slowdown pitch as fear mongering seizes the 'open AI' moral high ground. The US-China narrative split on AI governance is now a competitive lever.