Anthropic CEO Dario Amodei calls for government to block dangerous AI models

VentureBeat

Amodei's essay proposes mandatory third-party testing for frontier models trained with over 10^25 FLOPs, with civil penalties tied to global revenue. He cites AI's rapid progress: models went from barely writing code to autonomously executing complex cyber attacks in four years.

Anthropic

See 14 more sources

Anthropic has long advocated for transparency requirements for frontier AI, because the risks...3 days agoDario Amodei

Many of these policy ideas have common-sense appeal across the political spectrum, and the sooner...3 days agoDario Amodei

In addition to transparency, I now believe frontier models should face mandatory third-party...3 days agoDario Amodei

The essay also covers what AI’s steep trajectory means for jobs and the economy, scientific...3 days agoDario Amodei

Anthropic CEO Says Government Should Be Able to Block New Models3 days agoSeth Fiegerman

Both Anthropic and OpenAI mention the possibilities of slowing AI development in their latest "what...5 days agoEthan Mollick

Anthropic Urges Industry Coordination to Allow for a ‘Pause’ in AI Development if Risks Grow5 days ago

Anthropic CEO Warns AI Is Getting Too Powerful—While Releasing Powerful AI3 days agoJason Nelson

The company building God wants a kill switch...4 days agoFireship

I didn’t expect this from Anthropic5 days agoTheo - t3․gg

Policy on the AI Exponential3 days agoyjp20Discuss

It's not just Anthropic anymore, OpenAI researchers are signaling support for a global AI pause5 days agoEchoOfOppenheimerDiscuss

Has anyone able to verify Amodei's warning that "AI could soon build itself"? We're talking about RSI (that's proto-AGI).7 days agosourdubDiscuss

AI is advancing at a pace our policymaking institutions were never built for—and the gap between...3 days agoAnthropic

VentureBeat·Carl Franzen

3 days ago