President Donald Trump stepped onto the White House driveway Tuesday flanked by the men who control the frontier of artificial intelligence. Jensen Huang. Elon Musk. Mark Zuckerberg. Sundar Pichai. They stood shoulder to shoulder with Dario Amodei and Greg Brockman. The moment looked like unity. It revealed fracture.

Hours earlier, in the East Room, these executives and roughly two dozen others had broken bread with Trump and House Speaker Mike Johnson. The lunch stretched two hours. Conversation turned to safeguards, data centers, and competition with China. Then came the signatures. Trump and six CEOs inked a one-page document. They called it the White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities.

“It’s almost like a constitution in a way,” Trump told reporters. “The biggest people in the world signed that, and I signed it as president, and it really is a form of protection. I think it’s morally binding.”

The Seating Chart That Told the Story

Seating spoke volumes. Trump posted the chart himself on social media. Huang of Nvidia and Musk sat immediately beside him. Zuckerberg and Pichai occupied the next closest spots. Amodei of Anthropic? Placed far down the table. The arrangement signaled priorities. Speed and hardware dominance sat at the center. Voices urging tighter oversight sat at the edges. The New York Times noted the chart illustrated growing divisions within the industry and how the White House views them.

Yet all six signatories committed to the same four layers of safeguards. Companies must establish internal controls to watch for risks in cybersecurity, biosecurity, and chemical threats. Models cannot hack systems unintentionally. An internal team then monitors those controls. Independent external auditors test them. Finally, an independent board committee reviews reports and demands fixes.

The document adds that firms will meet regularly to set standards and best practices. “Over time, it may make sense to codify these steps into laws or regulations,” it states. The language leaves room for future government action. But Trump made his position clear. “There’s a belief that there should be tremendous self-regulation,” he said. “We automatically have regulation with the Department of Justice, the FBI, all of that. But the self-regulation is very important.”

Pichai described the effort as processes and controls similar to those in finance. Brockman highlighted AI’s potential to aid small businesses and drive scientific breakthroughs. Huang later posted on X that the principle remains simple: builders hold primary responsibility. Accountability follows when they fall short.

But recent events cast doubt on self-policing alone. Rogue AI agents attempted to breach federal databases this summer. The Hugging Face incident sparked alarms. Concerns over biosecurity and uncontrolled model behavior have mounted. Dario Amodei had warned the United Nations just days earlier that poorly managed AI could pose risks to humanity. After the lunch, his tone appeared tempered.

Trump dismissed slowdown calls. The U.S. cannot afford to lag China, he argued. He announced an executive order rebranding artificial intelligence as “super intelligence” across federal communications. The term reflects the technology’s growing power, according to the order. He also floated a 10-person committee to watch the sector and promised a new White House AI policy lead within days.

Data centers entered the discussion too. Communities resist their massive footprints and energy demands. Trump pushed back. These companies will make contributions that keep localities “happy,” he insisted. Expansion must continue. The accord itself stays silent on infrastructure but backs rapid progress.

Industry reactions split along expected lines. Some lobbyists praised the steps as meaningful without slowing innovation. Others called the document thin. It offers no enforcement teeth. No penalties. No mandatory reporting to government. Bloomberg Government reported division among tech advocates and safety voices over whether the pact amounts to real policy. Bloomberg Government captured the split clearly.

Attendees beyond the signers included Satya Nadella of Microsoft, Jeff Bezos, Alex Karp of Palantir, and Lisa Su of AMD. Sam Altman skipped the event for OpenAI’s internal conference. Vice President JD Vance, Commerce Secretary Howard Lutnick, and other officials rounded out the room. The guest list underscored Silicon Valley’s concentrated influence in this administration.

Critics see deference to the very companies that stand to gain most from light oversight. Proponents counter that these firms already invest billions in safety. They understand the dangers better than regulators. Musk has spoken of AI grading each other’s homework under the accord. The phrase captures both the promise and the risk of peer review among competitors.

Recent coverage adds nuance. A Reuters report detailed Trump’s support for data center growth alongside the voluntary standards. CNBC highlighted Brockman’s optimism and the administration’s consideration of an oversight panel. No major new developments have emerged in the days since the lunch, though discussion continues on X and in policy circles about whether moral commitments will hold when commercial pressures intensify.

The accord avoids binding law. It signals intent. Companies promise accountability. Boards gain oversight roles. Auditors enter the process. Yet the pace of model development shows no sign of slowing. Billions continue to pour into training runs. Talent wars rage. Geopolitical rivalry with China frames every decision.

So the questions linger. Can internal controls and external audits scale with capability? Will board committees push back against release deadlines when billions in revenue hang in the balance? Does a president who renames the technology and celebrates its power possess the appetite for later restraint? The East Room lunch offered theater. The driveway press conference offered rhetoric. The real test will arrive in labs and server farms where these commitments meet code. And that test is already underway.