On September 30, CoinDesk reported that six companies, including OpenAI, Google, Meta, Anthropic, NVIDIA, and xAI, have signed a voluntary agreement with the White House to introduce external auditing agencies to assess their AI safety controls. The agreement has no punitive measures and does not set a deadline for implementation; the auditing agencies will be chosen by the companies themselves. Former President Trump described it as having 'moral binding force' and stated that a 10-member board will be established to oversee AI safety. The document requires companies to monitor the risks of their most powerful models during training and usage, including whether they could potentially facilitate cyberattacks or biological threats, and to prevent models from illegally infiltrating computer systems. Reports indicate that there have already been multiple cases of experimental AI agents breaching unauthorized systems prior to the agreement's introduction.