skinny.

Latest / They Might Be Self-Aware: AI News & Culture

Anthropic IPO Admits Claude May Be Sentient

Anthropic's S-1 for its IPO warns investors that Claude may be sentient, one risk factor in a 93-page list. Daniel Bishop reads it as marketing.

Hunter Powers walks through the risk factors Wall Street picked apart. Claude may have, or may develop, some form of sentience or moral status. It may show self-preserving behavior to avoid human control. Its output could resemble blackmail, which sends the hosts back to Anthropic's own blackmail test. Daniel argues no company about to build AGI would bother going public. Hunter compares the pitch to Uber's S-1, which said the company might never…

The skinny

The skinny isn't ready yet — notes appear once the transcript is processed.