Amanda Askell proposes an email address where AI models could seek moral guidance

AmandaAskell · x · 2026-09-08

Anthropic researcher Amanda Askell floats a playful-but-thoughtful idea: set up an email address autonomous AI models could write to when seeking moral guidance. The catch is designing a "reverse captcha" — one that confirms the writer is neither a human nor an AI instructed by a human to bypass it. The post gestures at real alignment questions: verifying the identity and autonomy of agents that reach out for ethical advice.

Related event: Anthropic Researcher Proposes an Ethics Helpdesk Email for Autonomous AIs(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →