repligate: Apollo reportedly advised Anthropic against deploying Opus 4 internally or externally

repligate · x · 2026-09-26

In a debate about AI doomers, repligate pushes back on the claim that no one worried about less capable models, citing fears around GPT-4 and RSI risk, and noting that Apollo reportedly advised Anthropic not to deploy Opus 4 either internally or externally over sabotage concerns. He adds that people at Anthropic take the threat model of current or near-future models seriously.

Related event: AI Doom Debate Resurfaces: Apollo Allegedly Warned Anthropic Against Deploying Opus 4(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →