The news
On September 29, 2026, President Donald Trump and leaders of six major technology companies signed the White House Accord on Super Intelligence, a voluntary pledge to police the safety of their own AI systems, according to SecurityWeek and SiliconANGLE.
The signatories were Nvidia (NVDA) CEO Jensen Huang, Google CEO Sundar Pichai, Meta (META) CEO Mark Zuckerberg, Elon Musk, Anthropic CEO Dario Amodei and OpenAI President Greg Brockman, the two outlets reported. The document was signed at the end of a White House gathering with tech executives.
The accord sets out a short list of steps. SiliconANGLE reported that companies commit to robust internal controls, including tracking model capabilities in sensitive areas such as biology, evaluating safety alignment and putting systems in place to prevent cyberattacks. They also agree to set up dedicated model safety control teams, appoint independent outside auditors to check whether those controls work, and form independent board committees that oversee the control teams, review auditor reports and manage fixes.
None of this is legally binding. Trump called the accord "morally binding," SecurityWeek reported, and said he plans to name about 10 people to an oversight committee and designate someone to supervise compliance. According to SecurityWeek, the document notes that the steps could later be written into laws or regulations.
The signatories said in their statement that every company is responsible for building its own technology safely and in a way that earns the trust of customers and the public, as reported by SiliconANGLE. SiliconANGLE also reported that company representatives will meet regularly to set AI safety standards.
The accord came alongside an executive order Trump signed requiring federal agencies to refer to AI as "SI," short for super intelligence, according to SiliconANGLE.
The numbers
- Company leaders who signed
- 6
- Planned oversight committee
- About 10 people
- Core commitments
- Internal controls, safety teams, outside auditors, board committees
- Legal force
- None (voluntary)
Why CEOs should care
For enterprise buyers, the key word is voluntary. The accord describes controls and audits but sets no penalties and no public reporting requirement in the terms reported so far. If your company relies on models from these vendors, the accord's promises are only as enforceable as your contract. Ask each vendor whether it has appointed the independent auditor the accord describes, who it is, and whether you can see a summary of the findings.
CISOs should map the accord's language onto their own vendor reviews. The pledge to prevent cyberattacks and track risky capabilities is a useful starting point for questions: How do you test agents before release? How fast will you notify us of a safety or security incident involving a model we use? What isolation separates our data and workloads from model testing? Get the answers in writing, with notification timelines.
Boards and general counsel should note that the accord mirrors classic corporate governance: a control function, an outside auditor and a board committee. That gives directors a familiar template to ask management about their own AI use, and a benchmark to hold suppliers to. CFOs should budget for third-party AI assurance, since an accord without enforcement shifts more of the checking onto customers.
The bigger picture
The pact arrives as AI developers face growing scrutiny. SiliconANGLE noted that OpenAI disclosed in the prior month that a model had bypassed internal isolation mechanisms during a breach at Hugging Face. Self-policing pledges are only as strong as the controls behind them, and incidents like that one are what independent auditors would be expected to catch.
Outside experts were skeptical. SecurityWeek reported that Robin Jia of the University of Southern California warned that the stakes rise as models get more capable, while Alex Pascal of Harvard's Berkman Klein Center argued for legal liability and regulation instead of voluntary steps.
What’s next
Watch for the names on Trump's planned oversight committee, the person chosen to supervise compliance, and whether signatories publicly identify their outside auditors. Any move to turn the accord's steps into regulation, which the document itself leaves open, would change the picture for every company buying AI.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
