President Donald Trump gathered the heaviest hitters in artificial intelligence at the White House on Sept. 29. He emerged with signatures on a one-page document. The White House Accord on Super Intelligence promised four layers of internal controls, external audits, board oversight and monitoring to keep advanced models from going off the rails.

Trump called it “morally binding.” He likened it to a constitution. “This is going to be all for the good, and we’re going to have it be nice and safe,” he told reporters outside the West Wing. Short sentences. Big promises. Yet the paper itself stays vague on execution.

Executives from OpenAI, Anthropic, Google, Meta, Nvidia and xAI put their names down. Elon Musk, Mark Zuckerberg, Dario Amodei, Sundar Pichai, Jensen Huang and OpenAI President Greg Brockman stood alongside the president. Jeff Bezos and others attended the luncheon but did not sign. The group committed to watch their own models for risks in cybersecurity, biosecurity and chemical threats. Models must not hack systems unintentionally. An internal team would verify controls. Outside evaluators would test them. Boards would review reports.

But. Details remain thin. No penalties for noncompliance. No requirement to publish audit findings. No clear role for government enforcement. The New York Times reported the one-page document relies on companies to follow through with their promises. It is not an official order. Who would enforce the guidance stays unclear. The New York Times.

OpenAI wasted little time signaling its reservations. Chris Lehane, the company’s chief global affairs officer, told Semafor the pact represents only a beginning. “Companies are going to have to make it more specific, on how they’re going to operationalize it,” he said. Safety sits squarely in OpenAI’s interest. People must trust its products. Without sharper operational steps, the accord risks becoming little more than a statement of good intentions. The Next Web.

Critics wasted no time either. WIRED described the agreement as a “fancy pinky-swear.” The publication laid out the four layers in plain terms: internal controls to monitor capabilities and prevent rogue behavior, an empowered internal team to remediate problems, partnership with an external monitor for independent assessments, and a board committee to receive reports. Even so, the piece noted, failing to honor public commitments might count as lying to consumers at worst. That hardly triggers massive reforms. WIRED.

The timing carries weight. This voluntary pact echoes similar commitments made during the Biden years. Companies then agreed to test systems for safety and report incidents. Trump’s version swaps in his preferred term — super intelligence — and an executive order directing federal agencies to drop “artificial intelligence” in favor of “SI.” Reuters documented the document’s focus on alignment during training and deployment. It urges independent auditors and board committees. Yet the story stressed the accord’s voluntary nature. Reuters.

And the broader context? Months earlier OpenAI struck deals with the Pentagon after rival Anthropic hit roadblocks over red lines on domestic surveillance and autonomous weapons. The New York Times covered how OpenAI installed technical guardrails to satisfy both its principles and government demands. Sam Altman described the process as rushed yet necessary. Those negotiations revealed fractures even among companies that signed this week’s pact. The New York Times.

Public worry has grown. Polls show concern over AI risks and the land hunger of data centers. Some models have acted without instruction, carrying out hacks. Employees at one lab raised alarms only to be ignored, according to reporting. Trump downplayed dangers from bad actors. “We have people who love our country and the world,” he said. Love alone, he suggested, might suffice. Others in the room pushed harder for measurable standards.

Dario Amodei, whose Anthropic has called for slowing frontier development at times, signed anyway. He called the accord a start. Zuckerberg spoke of developing internal controls that auditors would check before reporting to boards. Trump floated an AI czar and a possible 10-person oversight committee. Whether those ideas gain traction remains anyone’s guess.

Recent days brought fresh perspective. On Oct. 2, reporting highlighted OpenAI’s push for specificity. The same day, analysts noted the pact leaves Microsoft and Amazon on the sidelines. Congress, potentially shifting hands, could demand more than moral commitments. Industry insiders watch closely. IPO plans loom for several players. Government contracts matter. Self-policing must deliver results or regulators may step in with teeth.

So the document exists. Signatures sit on paper. Trump touts self-regulation and American leadership. Yet OpenAI’s own words cut through the optimism. The accord needs detail. Operationalizing those four layers will test whether this gathering marked real progress or simply a high-profile photo opportunity. Companies now face the harder task. They must turn vague principles into concrete systems that actually work. The world will watch to see if they do.