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

As major AI laboratories increasingly move to establish internal auditing frameworks to address safety concerns and the risks posed by rogue agents, industry critics are questioning the efficacy of these self-policing measures. The article suggests that while internal oversight is a positive step, it may be insufficient if the foundational security protocols—the 'front door'—remain porous. Instead of relying solely on internal audits, the piece argues that labs should prioritize hardening their core infrastructure and access controls to prevent unauthorized exploitation. By focusing on fundamental security hygiene rather than just reactive auditing, companies could more effectively mitigate the dangers associated with advanced artificial intelligence. The discussion highlights a growing tension between the rapid pace of AI development and the necessity for robust, transparent safety standards that go beyond internal corporate governance.