AI Slowdown Divide: Anthropic Safety Plan Faces Nvidia Pushback
Anthropic CEO Dario Amodei proposes frontier AI safeguards as Nvidia CEO Jensen Huang and Donald Trump reject slowing development amid enforcement doubts.
Summary
As of September 20, 2026, Anthropic CEO Dario Amodei is urging frontier AI labs to “pace the frontier,” an approach quickly supported by OpenAI CEO Sam Altman and, to a degree, Elon Musk. Proposals include independent third party evaluators inside labs such as Anthropic and OpenAI to monitor safety practices and incidents, common standards and limits among major AI companies in democratic countries, and international coordination. The framework leaves dangers, enforcement and the meaning of pacing undefined, while none of its measures inherently requires slower development. Anthony Ha viewed the emerging consensus as encouraging but questioned whether it represented a genuine slowdown.
Nvidia CEO Jensen Huang opposes slowing development and echoed President Donald Trump’s claims that the AI backlash is a hoax and regulation unnecessary during an onstage call at the All-In Summit. Kirsten Korosec warned that coordination among frontier labs could begin to resemble a cartel. Sean O’Kane argued that existing regulation and market pressure offer weak safeguards because federal enforcement appears unlikely, enterprise customers have limited leverage, companies are unlikely to abandon OpenAI’s Codex for Anthropic’s Claude Code over principle, and abundant investment can help labs absorb losses.
Positives
- Dario Amodei’s proposal quickly gained support from OpenAI CEO Sam Altman and, to a degree, Elon Musk.
- Independent third party evaluators could monitor safety practices and incidents inside frontier labs including Anthropic and OpenAI.
- Major AI companies in democratic countries could coordinate on common safety standards and limits.
- International coordination could extend frontier AI safeguards beyond individual companies and national markets.
Risks & concerns
- Amodei’s framework does not define the relevant dangers, enforcement mechanisms or practical meaning of pacing the frontier.
- Jensen Huang and Donald Trump reject a slowdown, with Nvidia benefiting from continued rapid AI development.
- Kirsten Korosec warned that coordination among frontier AI labs could develop cartel-like characteristics.
- Sean O’Kane said weak federal enforcement and limited consumer choice undermine existing regulatory and market safeguards.
- Enterprise customers may not switch from OpenAI Codex to Anthropic Claude Code over principle, reducing commercial accountability.
- Heavy investment could let AI labs absorb customer losses without changing unsafe practices.
