The debate about pacing frontier AI has settled on a practical point: safety commitments mean little unless someone independent can check them. But checking things properly takes people who know what to check against, and standards to check against in the first place. Some of that groundwork already exists. ISO/IEC 42001 sets out how an organisation manages AI responsibly. ISO/IEC 42006 defines what a body needs to audit and certify against it, including competence and impartiality. ETSI EN 304 223 sets baseline security requirements across the AI lifecycle. AIQI Consortium, which we co-founded with UKAS, has developed free self-paced courses on all three, with input from accreditation bodies across several countries. Open to anyone who finds them useful. #TICCouncil #AI #AIQI
For all the recent noise and despite how politicised and heated the once-boring topic of AI auditing has become, it’s a great reminder that turning to balanced and carefully-written standards like ISO 42001 is still the leading solution.
Kamilla Harcej
1mo
Exactly. This is precisely what we are addressing through TauDIL GOS™ - the proprietary, IP-protected Governance Operating System owned by TauGuard Limited. Our architecture is designed to make governance an enforceable part of execution, not simply an advisory function. Governance must have the structural authority to prevent unauthorised actions before they happen.
Peter Hill
1mo
Agreed. Safety commitments mean little unless someone independent can check them. These independent audits are best performed by internal auditors.
🔗 https://www.aiqi.org/