On Tuesday, President Donald Trump gathered the most powerful technology executives on the planet at the White House, stared down the existential risks of artificial intelligence, and emerged with a deal he described as 'morally binding.' It is not legally binding. It is not regulation. It is, in the technical policy sense, a promise.
What They Actually Signed
According to NPR, Trump and the CEOs of Anthropic, Google, Meta, OpenAI, Nvidia, and xAI all put their names to a voluntary accord that Trump posted on Truth Social Tuesday evening. Signatories include Dario Amodei, Sundar Pichai, Mark Zuckerberg, Greg Brockman, Jensen Huang, and Elon Musk, who apparently took a break from dismantling the federal government to sit on Trump's left at a White House luncheon.
The accord outlines four voluntary steps: companies will implement 'robust internal controls,' partner with an 'independent external auditor,' establish a board committee to review audit reports, and, at some unspecified point in the future, maybe consider laws. 'Over time, it may make sense to codify these steps into laws and regulations,' the accord says. Bold stuff.
Trump told reporters the agreement would be 'morally binding' and that roughly ten people would be named to a committee to 'watch over the whole enterprise.' He said he'd name someone to oversee the accord in the coming days, after consulting with the industry it is supposed to oversee. The fox is helping design the henhouse inspection program.
The Part Where Experts Point Out the Obvious
Robin Jia, an assistant professor of computer science at the University of Southern California, told NPR he's skeptical of putting 'so much faith in self-policing.' Jia pointed to recent AI hacking disclosures and noted that as the models get more powerful, the consequences of those failures get worse. This seems like a reasonable thing to be worried about.
Alex Pascal, executive director of the Berkman Klein Center for Internet and Society, was more direct. 'The only way to reduce the risks and harms of AI to Americans,' Pascal told NPR, 'is through robust legal liability, regulation and fundamentally changing the race dynamics that are driving unsafe decision-making by frontier AI developers whose costs are borne by the American people.' That is a polite academic's way of saying a pinky swear is not going to cut it.
Even Dario Amodei, CEO of Anthropic and one of the louder voices in the AI industry calling for a slowdown, told reporters Tuesday that 'the technology has very real risks' and that 'the mechanism, how we address those risks is still under discussion.' That last clause is doing a lot of heavy lifting. The mechanism is still under discussion. The technology, meanwhile, is not under discussion. It is being built at a speed that has never been seen before in human history.
Trump's Position, Stated Without Ambiguity
Earlier in the day, before the accord signing, Trump made his actual governing philosophy crystal clear. 'I will never stifle the growth of a technology that will be bigger than the industrial revolution,' he told the crowd at the unveiling of America.gov, a new AI government chatbot intended to help Americans navigate federal services. 'We can't stifle it, and we're leading by a lot.'
He also said the U.S. government would not support calls for 'guardrails' or other restrictions on AI development. So to recap the sequence of events: Trump says there will be no guardrails, then signs a voluntary document that functions as guardrails on the honor system, and calls it a win. House Speaker Mike Johnson was also present for this, nodding along.
The America.gov chatbot rollout happened at a daylong White House event where multiple Trump administration officials celebrated what Trump insists on calling a 'Golden Age' powered by what he insists on calling 'super intelligence' rather than artificial intelligence. Secretary of State Marco Rubio highlighted that the chatbot could eventually let Americans apply for a passport fully online. That is a genuinely useful thing, to be fair. It exists alongside the part where the most powerful AI companies in the world agreed to audit themselves.
Nothing Here Is Actually New
NPR's reporting includes a telling detail that deserves its own paragraph: some of the actions outlined in the accord are measures these companies are already taking in some form, or have previously committed to doing. So the historic White House AI summit produced a document in which tech giants pledged to continue doing things they were already doing, voluntarily, with no legal consequence if they stop.
Shri Narayanan, a professor of electrical and computer engineering at the University of Southern California, offered the most charitable read to NPR, saying the accord balances innovation with safety considerations and that the safety work includes improving model security. That's true as far as it goes. The question is whether a voluntary accord signed at a lunch where Elon Musk sat next to the president constitutes a meaningful check on an industry spending hundreds of billions of dollars to build technology that its own leaders describe as potentially world-altering.
On the data center front, Trump told reporters that the tech executives agreed to offset local community opposition to massive data centers by providing more financial support to local schools and reducing energy costs for nearby residents. The bipartisan opposition to data centers in local communities is real, and it's growing. Trump's solution is for the companies building them to offer some goodwill gestures. Voluntarily.
The Dingo Take
You are supposed to look at this accord and feel reassured. You are supposed to read 'robust internal controls' and 'independent external auditor' and feel like someone is in charge. You are supposed to trust that Elon Musk, sitting to the president's left at a White House luncheon, has signed a document that will meaningfully constrain his own AI company's behavior in ways that benefit you rather than him.
This is the same regulatory philosophy the Trump administration has applied to every other industry: let the companies decide what's safe, ask them to write it down, call it a deal, hold a press conference. It worked great with financial products in 2007. It's going great with social media. The AI industry is, by the admission of its own CEOs, building something potentially more consequential than either of those things, and the governing response is a Truth Social post and a promise to name an oversight person sometime soon.
Dario Amodei said the risks are very real and the mechanism for addressing them is still under discussion. Jensen Huang ate lunch next to the president. The accord says laws might make sense eventually. Meanwhile the data centers are going up, the models are getting better, the auditors will be chosen by the companies being audited, and nobody who signed that piece of paper will face any legal consequence if they ignore every word of it. But it's morally binding. Sleep tight.
Comments