Trump signs 'morally binding' AI accord as Brussels enforces its own rulebook
President Donald Trump said on Tuesday that he and tech executives had signed a "morally binding" AI document at the White House. The EU's AI Act moved into enforcement the same week, leaving two rival rulebooks in force.

President Donald Trump said on Tuesday that he had signed a "morally binding" artificial intelligence document with tech leaders. The signing followed a luncheon at the White House, according to a CNBC account of the event. House Speaker Mike Johnson, standing beside him, called the same text a statement of principles that is "voluntary on behalf of the industry." That is a two-word gap between the president and the speaker. It is the whole of US AI policy in miniature.
Trump said the administration was considering a 10-person committee to oversee the AI industry. He said he would name a new AI czar within three to four days. He also described large data centres as "a very positive thing." Tech leaders would "make communities very, very happy" or be "forced to go overseas," he said.
Two rulebooks, one industry
Outside the White House, Anthropic chief executive Dario Amodei said rules addressing AI risks were "still under discussion." AMD's Lisa Su said she left "very encouraged." Neither statement is a compliance document. Meanwhile, in Brussels, the EU AI Act has moved into its enforcement phase. The trade press covered the move this week, and it led the headline list, including HR Executive's 28 September report on cross-border HR content and App Developer Magazine's note on the global implications for American vendors.
Any company selling models or cloud capacity into Europe faces a practical problem. It is not the American accord's wording. It is that the accord has no wording that a regulator can enforce, while the EU text does. A vendor cannot file a "morally binding" statement. It can only file a conformity assessment, a technical file, or nothing. The timing is not incidental.
OpenAI said on Monday it had decided not to release its GPT-6.1 Astra model. The company concluded it "didn't quite meet the bar in terms of staying within scope and authorization," according to Saachi Jain, the company's head of safety systems. The Wall Street Journal was first to report the decision, and CNBC confirmed it. OpenAI had released the earlier GPT-6 Astra earlier in September. It added GPT-6 Sol and GPT-6 Luna tiers last week.
Self-regulation meets a filing deadline
Trump's answer to that is self-regulation. "There's a belief that there should be tremendous self-regulation," he told reporters. He added that the Department of Justice and the FBI already provide "regulation." That is not a framework; it is a statement of faith. It collides with a European regime that asks for documents rather than beliefs.
"It does not appear that a customer can buy a new car that does not track you." Sarah Elizabeth Gillespie, Northeastern University, in a Consumer Reports study
The enforcement gap is already visible in adjacent markets. A Northeastern University study published with Consumer Reports tested 21 vehicles and their companion apps. The test ran in a Faraday tent at CR's Connecticut test centre. It found driver data flowing to Amazon, Google, Meta, Microsoft, Pinterest, Snap and Yahoo. Almost a quarter of the vehicle apps sent personally identifiable information, including names, VINs and precise locations. The study will be published this week. No US regulator has moved on that, and no European one needs a new law to do so.
Capital markets are behaving as if spending, not rulemaking, will settle the regulatory question. Torsten Slok, chief economist at Apollo, published a note on Tuesday. He pointed out that Wall Street's sector analysts are internally inconsistent. Tech analysts expect the sector's operating cash flow to more than double to roughly $2.4 trillion by 2028, an increase of over $1.2 trillion. Analysts covering tech's own customers expect far less. "Both cannot be right at the same time," Slok wrote.
If the customer-side forecasts are closer to reality, the AI buildout is being financed against demand that has not yet been contracted. If the tech-side forecasts are right, the customers are about to spend a great deal more than their own analysts expect. Either way, someone is writing a very large cheque. No one has yet established which regulator will be watching when it clears.
The enforcement question nobody answered
The EU's position has been stated repeatedly by its own officials. The tech chief said AI guardrails are "inevitable" for the United States, as reported on 4 September. That is a political prediction, not a legal mechanism. The mechanism is the Act itself. It applies to any provider placing a system on the EU market, regardless of where the provider is incorporated.
Google is already testing that boundary. The company has challenged EU orders to open Android and search data to rival AI services. chinadailyhk reported the fight on 28 September and Crypto Briefing reported it on 29 September. That dispute is about access and interoperability, not about safety claims. It is a better guide to what enforcement will actually look like than any summit communique.
The American side, meanwhile, is moving in the opposite direction. Florida has asked a court to block OpenAI from developing new models, Politico reported. The state separately aims to stop ChatGPT from behaving like a person, according to The Verge. Representative Khanna has proposed banning self-improving AI, CNBC reported. Senator Cruz blocked a bill that would have given regulators 45 days to review AI models before launch, according to Pasquale Pillitteri's 29 September writeup.
So the United States has a voluntary accord, a proposed 10-person committee and a pending AI czar. It also has at least two state-level actions, one blocked federal bill and one proposed federal ban. Europe has a statute that entered enforcement this week. Those are not comparable instruments. Treating them as a transatlantic "dialogue" flatters the weaker one.
What the accord does not say
The CNBC report does not quote any text from the document Trump signed. It describes the event, the attendees and the statements made around it. Amodei, Nvidia's Jensen Huang and Tesla's Elon Musk were among those present. That is the full extent of what is publicly known about the accord's contents.
Anthropic, for its part, has filed for an IPO. That filing warns that AI could threaten humanity, according to the Financial Times. The company is making two different claims to two different audiences. It makes one argument to public-market investors while signing a non-binding White House statement. Both are on the record.
The enforcement asymmetry has a commercial edge. Comcast said on Tuesday it is deploying fibre sensing technology. The system analyses vibrations in fibre to detect construction near buried lines. Future plans could include dispatching drones to inspect damage, vandalism or wildfires. Spectrum chief executive Chris Winfrey spoke at SCTE TechExpo. He said multi-gigabit deployment "opens up the door" to deeper conversations with global Big Tech providers. He said cable operators want products built to multi-gig capability across the US footprint. Both are infrastructure bets made without reference to which AI rulebook will govern the services running on top.
GM's Ultium Cells joint venture with LG Energy Solution said on Tuesday that its Spring Hill, Tennessee plant will be "the first in the world to mass produce prismatic LMR battery cells." The cells have 33% higher energy density than LFP at comparable cost. Upgrades start later this year, with completion by 2028. The plan adds 500 jobs, taking the site to 1,700. Alibaba, for its part, became official technology and cloud partner and jersey patch sponsor of the Brooklyn Nets. The multi-year deal was announced on Monday. It powers a 360-degree replay system at Barclays Centre from the 2026-27 season.
None of that waits for Brussels. That is the point. The EU AI Act's enforcement phase is now the only binding AI rulebook with global reach that is actually operating. It is operating while the largest AI market in the world relies on a document whose signatories describe it in two different ways. Companies will comply with the version that has penalties attached. They will describe the other one as a statement of principles.
The gap will not be closed by another luncheon. It will be closed by whichever regime first fines someone large enough to make the evening news.
Sources
13- 01Trump says he and tech leaders signed AI agreement that is 'morally binding'EN
- 02OpenAI abandons plan to release upcoming model as safety concerns escalateEN
- 03The Download: climate tech companies to watch and AI's discovery problemEN
- 04GM's new EV battery tech will cut costs without sacrificing performance or rangeEN
- 05Your Car Is Sharing Data With Big Tech Companies, Study FindsEN
- 06Tech's Trillion-Dollar Internal InconsistencyEN
- 07Comcast deploys fiber sensing technologyEN
- 08Multi-gig at scale will help cable engage with Big Tech – Spectrum CEOEN
- 09Alibaba deepens sports tech push with new Brooklyn Nets partnershipEN
- 10Beam sues YolTech and Serapha Bio for allegedly stealing its gene editing techEN
- 11These Tech Workers Made ChatGPT Drive a Toyota CorollaEN
- 12The Tech Industry: An AutopsyEN
- 13The Book Your Tech Boss Doesn't Want You to ReadEN
All figures and quotations in this text come from the sources listed below.
Content prepared by the editorial team with AI assistance.
Comments
0- No comments yet — be the first.