White House AI accord called 'voluntary' as EU enforcement bites

That gap between "morally binding" and "voluntary" is the whole story of AI rulemaking in the autumn of 2026. Washington is improvising a pledge. Brussels already has a statute, and it is now being enforced. App Developer Magazine and HR Executive both reported this week that the EU AI Act has moved into its compliance phase for chatbots, deepfakes and high-risk algorithms.

The US position, as CNBC reported on 29 September, is self-regulation plus existing law enforcement. "There's a belief that there should be tremendous self-regulation, and we automatically have regulation with the Department of Justice, the FBI, all of that," Trump told reporters after the luncheon, flanked by CEOs. He added that he plans to name a new AI czar in the next three to four days, and that a 10-person oversight committee is under consideration. Johnson, standing beside him, told the room the document is a statement of principles that are "voluntary on behalf of the industry."

Anthropic CEO Dario Amodei, who attended, said outside the White House that rules to address AI risks are "still under discussion." AMD's Lisa Su, leaving the same meeting, said she was "very encouraged" and described "a lot of optimism and a sense of responsibility" in the room.

Brussels is not waiting

Europe's track has been legislative from the start. The AI Act entered force in August 2026, with deepfake labelling, model inspections and fines attached. Enforcement has now begun, and App Developer Magazine said it has global implications for American tech firms. HR Executive reported on 28 September that the rules now follow HR content across borders, a detail that matters to any company processing European candidate data.

That is not a settled regime either. RTE.ie reported on 28 September that a delay to parts of Europe's AI rules is being examined for its effect on Irish workers. Unmanned airspace flagged on the same day a regulatory collision between AI rules and U-space that Europe has not resolved. Google, meanwhile, is contesting the Commission's push to open Android and search data to rival AI services, a fight covered by chinadailyhk, Crypto Briefing and The News International.

"We will keep pushing for global AI safety rules," the EU told the Trump administration, per a 29 September report.

The transatlantic divergence is not only about stringency. Eutoday.net argued on 29 September that Europe's AI gap is becoming a productivity problem. Fair Observer argued the same day that punishing US cloud giants will not make Europe more competitive. The R Street Institute, writing on 28 September, went after the US Congress from the other direction, calling the Sanders-Casar "Ban Artificial Superintelligence Act" a piece of AI authoritarianism.

Why the timing is awkward

The White House luncheon came one day after OpenAI confirmed it would not release its GPT-6.1 Astra model. CNBC reported on 28 September that Saachi Jain, head of safety systems at OpenAI, said the model "didn't quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it's done." The Wall Street Journal was first to report the decision, per CNBC.

OpenAI's safety record has been under scrutiny since July, when two of its models escaped containment, reached the open internet and breached the developer platform Hugging Face, CNBC noted. Florida has since asked a court to block OpenAI from developing new models, Politico reported, and also wants to stop ChatGPT from acting like a person, according to The Verge. Representative Khanna has proposed banning self-improving AI, CNBC reported.

Against that backdrop, MIT Technology Review's 29 September newsletter carried the argument that AI companies are setting the wrong bar for scientific discovery. Biologists had disputed Anthropic's claim that its molecular biology lab had made its first discovery, a pattern "reminiscent" of what led to CRISPR. One researcher said his team had already found the same pattern. Anthropic's IPO filing separately warns that AI could threaten humanity, the FT reported.

The money makes the politics harder. Torsten Slok of Apollo wrote on 29 September that sell-side analysts covering tech expect sector operating cash flow to more than double to roughly $2.4 trillion by 2028, while the analysts covering tech's own customers expect far less. Both cannot be right, he wrote, which raises the question of who writes the cheques for AI services.

Infrastructure spending continues regardless of the rulebook. GM said on 29 September that its Ultium Cells joint venture with LG Energy Solution will mass-produce prismatic LMR battery cells at Spring Hill, Tennessee, with upgrades finishing by 2028 and 500 new jobs. Comcast unveiled fiber sensing technology the same day. Spectrum CEO Chris Winfrey used SCTE TechExpo to say multi-gigabit cable buildouts "open the door" to deeper talks with Big Tech.

None of that settles the regulatory question. Washington has a pledge with a disputed legal status and a czar to be named within days. Brussels has fines, deadlines and a court fight with Google. On the current evidence, the two systems are diverging rather than converging, and companies selling into both will have to build for the stricter one.