Amodei Outlines Three Plans to Slow AI Development — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Dario Amodei published a blog post calling to 'pace the frontier' of AI development, citing two reasons for acting now: the OpenAI-HuggingFace hack and AI's accelerating ability to build the next generation of AI models
- Anthropic is unilaterally committing to embedded third-party evaluators from organizations like METR, who would receive company badges, desks, and laptops, plus access 'mostly comparable to what internal risk assessment teams have'
- Sam Altman said OpenAI will follow Anthropic's lead on embedded evaluators, calling it a 'good idea,' while Elon Musk posted 'Dario is right' on X
- Amodei's second proposal calls for AI companies in democratic nations to coordinate on common safety standards, with Amodei requesting that the US government mediate and issue narrow antitrust waivers for safety conversations
- The third proposal calls for global coordination including with authoritarian governments like China, with potential starting points being narrow prohibitions such as banning AI-enabled biological weapons production
- Amodei addressed the China counter-argument by arguing that refusing to sell powerful chips and semiconductor manufacturing equipment to Chinese companies, plus cracking down on model distillation, could widen America's lead 'significantly over the next 3–5 years'
- The post follows Anthropic researcher Jacob Coxon's resignation, in which he warned that leading AI companies are 'gambling with our lives' and that people building the technology 'earnestly believe it could kill us all by the end of the decade'
Why it matters: Anthropic's unilateral commitment to embedded third-party evaluators — and OpenAI's pledge to match — creates immediate structural pressure: with two frontier labs voluntarily opening themselves to outside auditors, the mechanism could become a de facto industry standard even without government mandate. Critics like Brian Merchant counter that such self-administered oversight amounts to 'regulatory capture' that would primarily serve Anthropic and OpenAI rather than the public.
Ask SkimNews
