Six AI companies sign White House audit agreement; critics say the loophole is that companies choose their own auditors.

PANews October 17 news: According to an opinion article published by CoinDesk, Samuel Vance-Law, research director at the Decentralization Research Center, pointed out that the White House frontier AI agreement signed on September 29 by Nvidia, OpenAI, Anthropic, Google, Meta, and xAI (which Trump called "morally binding") has key flaws: the agreement does not specify who selects the auditors, what content can be reviewed, whether the conclusions will be made public, and there are no penalties for breach; in practice, each company selects its own auditors and decides the scope of review. He cited the 2000 case in which Enron hired Arthur Andersen for auditing, which later led to the Sarbanes-Oxley Act, as a cautionary precedent, and cited an assessment showing that the average compliance rate of 16 companies with the 2023 White House voluntary commitments was only 53%, with 8 companies scoring zero on the item of assisting external error detection. He argued that auditors should be able to independently choose what to test, companies should disclose the auditors and the scope of review, and oversight should be conducted by an institution independent of the companies.