White House Accord on AI Audits Lacks Clear Rules for Independent Oversight
Leaders from major tech companies, including Google, OpenAI, Anthropic, Meta, xAI, and Nvidia, signed the White House Accord on Super Intelligence on September 29. This accord includes a commitment to engage independent outside auditors to verify the effectiveness of their artificial intelligence safety controls. While these commitments are currently voluntary, there is a possibility they could be codified into law. However, a significant concern highlighted by an auditor with over 25 years of experience is the absence of clear rules governing these independent audits. The accord does not define what constitutes an 'independent' auditor, the extent of access auditors should have, or the standards against which they should measure AI systems. This lack of defined parameters means that two companies could claim to have passed an independent audit, yet the rigor and scope of those audits could differ significantly, potentially undermining public trust and the credibility of the audit process.
