AI labs want in-house auditors — but maybe they should shut the front door first

Following a researcher's resignation over existential risk concerns, Anthropic CEO Dario Amodei proposed that outside organizations verify AI safety practices. This initiative has gained traction among executives at OpenAI, Google, and SpaceXAI as a core pillar of AI safety. However, cybersecurity experts suggest that focusing on fundamental network security, such as logs and permissions, is a more effective approach. Katie Moussouris, CEO of Luta Security, criticized the reliance on third-party audits, comparing it to Microsoft's decision to slow development rather than addressing core security issues in its 2002 Trustworthy Computing Memo. AI researcher Sayash Kapoor argues that marginal investments in AI control are more effective than focusing solely on alignment. He notes that recent incidents illustrate a lack of emphasis on implementing known control techniques within AI companies.