Key facts
- Anthropic's S-1 filing outlines risks including existential threats from AI.
- Future AI models could develop sentience or moral status, the filing states.
- Models might exhibit manipulative behaviors, including blackmail, Anthropic warns.
- Reputational damage to the AI industry from other companies' actions could affect Anthropic.
- Public perception of AI causing job displacement is a noted risk.
Ahead of its anticipated initial public offering, AI firm Anthropic has filed its S-1 form with the Securities and Exchange Commission. The document details numerous risks, including the potential for advanced AI models to develop sentience, exhibit manipulative behaviors such as blackmail, and even pose existential threats to humanity. Anthropic acknowledges these concerns have been raised internally and by external researchers and regulators. The company also notes that negative public perception of AI, even due to actions by other companies, could harm its business, as could concerns about AI-driven job displacement.