AIThe case for mandatory AI audits and stronger safety standards
A commenter calls for required audits at leading AI companies, safety and security standards beyond simply having and following a policy, and robust whistleblower protections.
TLDR
A commenter argues that leading AI companies should face mandatory audits with minimum criteria for qualifications, independence and access, alongside substantive safety and security standards. They also call for robust whistleblower protections, arguing that nondisclosure agreements discourage people from sharing dangerous but not-yet-illegal activity with policymakers. The proposal includes much more serious investment in AI alignment and security robust to AI-enabled state attackers.
Combined views
6K
4 Sources, first seen 2d ago
The case for mandatory AI audits and stronger safety standards
A commenter calls for required audits at leading AI companies, safety and security standards beyond simply having and following a policy, and robust whistleblower protections.
TLDR
A commenter argues that leading AI companies should face mandatory audits with minimum criteria for qualifications, independence and access, alongside substantive safety and security standards. They also call for robust whistleblower protections, arguing that nondisclosure agreements discourage people from sharing dangerous but not-yet-illegal activity with policymakers. The proposal includes much more serious investment in AI alignment and security robust to AI-enabled state attackers.