A Voluntary AI Safety Pledge Gets Teeth When Auditors Report to the Board
Sacks argues no new law is needed, because audit reports create a legal duty for directors to act.
All-In with Chamath, Jason, Sacks & Friedberg · Trump's Super Intelligence Summit, AI Safety Accord, GDP Beats, Midterm PredictionsSacks says a voluntary safety accord signed by six leading AI companies is meant to work like a requirement. The mechanism is audit reports sent to the board, with no new law needed.
David Sacks, who was in the White House meeting, said the companies developing the most advanced models agree that they are the responsible actors, and cannot shift blame to the models themselves or to someone else like the United Nations.
Each company also agrees to set up internal safety controls, and to have a separate internal team check that those controls are actually working.
Independent auditors then come in to "validate and verify that the internal controls are working," Sacks said. Sacks said the most effective part of the accord is these external audits, and said in his view they should be done by professional accounting firms like EY, PwC, or KPMG, not by advocacy groups. Investor Chamath Palihapitiya said "you're gonna see EY in short order here announce something."
Finally, each company forms an independent committee of its board to receive the audit reports. That delivery is what gives the system force.
As investor Gavin Baker pointed out in a discussion cited by Sacks, a board cannot disregard findings delivered by an outside reviewer. Directors have a fiduciary duty, meaning a legal obligation to act in the best interests of the company and oversee major risks. If they ignore a warning, Sacks said, they risk losing their D&O insurance, the directors and officers liability insurance that protects board members if something goes wrong.
That board-plus-auditor structure was new. Sacks said the White House had first circulated a two paragraph statement focused on responsibility and a safety working group to share best practices. Sacks said Mark Zuckerberg, who runs Meta, and Jensen Huang, who runs Nvidia, came forward with the idea. He thinks the rest of the room agreed.
The result, Sacks argued, is a voluntary deal that becomes, in his words, de facto required. Sacks said the Federal Trade Commission and the Securities and Exchange Commission have the ability "to enforce the public commitments of these companies," and said the controls can be put in place quickly, without Congress passing a law or countries negotiating a treaty.
Questions people ask
What is a fiduciary duty for a corporate director?
It is the legal obligation of directors to act carefully, honestly, and in the best interests of the company and its shareholders. In practice, it means they must pay attention to known risks, ask questions, and make sure management addresses them. If they ignore a clear warning, shareholders and courts can hold them personally responsible.
What does D&O insurance do?
Directors and officers liability insurance pays legal costs and damages if board members are sued over their decisions. Insurers expect boards to follow standard oversight practices, like reviewing outside audits. If directors ignore audit findings, insurers can limit coverage or refuse to renew the policy, leaving directors exposed.
How could the FTC and SEC enforce a voluntary AI pledge?
Both agencies police what companies tell the public. If a company publicly commits to safety audits and board oversight and then fails to do it, the FTC can bring a deceptive practices case and the SEC can allege misleading disclosures to investors. That existing authority means regulators do not need a new AI safety law to act.