Source: TechCrunch
Frontier labs like OpenAI, Anthropic, and Google have spent years discussing AI safety in the abstract while keeping their actual containment protocols secret. Regulators, investors, and the public have no way to verify whether the industry's self-policing works. The absence of documented procedures suggests either that containment is harder than labs admit, or that they're unwilling to reveal vulnerabilities that might undermine investor confidence or invite regulatory scrutiny. The opacity makes it difficult to assess whether leading labs can manage AI risk responsibly.