Zuckerberg, Huang and Musk Reportedly Lobbied Trump to Block Demis Hassabis’s AI Safety Regulator Plan

A proposed US AI safety body, championed by Google DeepMind co-founder Demis Hassabis and modelled on Wall Street’s self-regulatory watchdog, has been stalled after three of the world’s most powerful technology executives reportedly lobbied President Donald Trump against it. The Wall Street Journal, citing people familiar with the matter, reports that Mark Zuckerberg of Meta, Jensen Huang of Nvidia and Elon Musk — who leads SpaceX, Tesla and xAI — each spoke separately with Trump and raised objections to the framework. The plan has since been shelved within the administration, though not formally abandoned.

The Hassabis proposal called for a “Frontier AI Standards Body” funded by the AI industry but operating under government oversight — not a government agency itself, but an independent organisation with technical experts conducting safety evaluations. The model was explicitly inspired by FINRA, the Financial Industry Regulatory Authority, which oversees US brokerage firms and exchange markets and currently regulates thousands of firms and millions of brokerage professionals. Under the draft framework Hassabis outlined in a governance essay published in mid-July, powerful AI models would be required to submit to safety evaluations up to 30 days before public release, with assessments focused on risks in areas such as cybersecurity, biosecurity and autonomous deception. The system would initially operate on a voluntary basis, with the potential for mandatory pre-release checks once the process was established.

How the Proposal Reached the White House

Hassabis moved the concept from a public essay into direct lobbying, briefing senior US officials including the Treasury Secretary and the Director of the White House Office of Science and Technology Policy. That it reached such levels suggested the framework had genuine momentum before the reported intervention by the three CEOs.

Anthropic’s leadership is also reported to have supported the Hassabis approach, highlighting a genuine split within the AI industry. Dario Amodei’s backing is notable given Anthropic’s reputation for prioritising safety research — and it places the opposing camp, Zuckerberg, Huang and Musk, in a rather different position on the question of oversight.

Why the Three CEOs Reportedly Objected

The objections raised by Zuckerberg, Huang and Musk, according to Forbes analysis and other coverage drawing on unnamed sources, centred on competitive fairness. Their concern was that a FINRA-style body could effectively entrench the leading frontier AI labs — OpenAI, Anthropic and Google DeepMind — as the dominant players, since those companies would be best positioned to shape and comply with new standards. For Nvidia, whose chips power most of the AI industry, and for Meta, which develops its own frontier models including the Llama series, a regulatory architecture designed around a handful of incumbents could create an uneven playing field.

That’s a legitimate concern, and it’s one that critics of industry self-regulation raise from the opposite direction too. Civil society groups and policy analysts have long warned that any self-regulatory body risks being shaped by whichever companies have the most influence over it — so the argument cuts both ways.

The Trump administration’s stated position on AI has consistently emphasised light-touch regulation and maintaining US competitiveness against China. David Sacks, the technology investor serving as Trump’s AI adviser, is reported to be among those pushing to keep formal regulatory burdens minimal. The stalling of the Hassabis proposal fits that pattern.

What Hassabis’s Framework Was Actually Trying to Do

It’s worth being clear about what the proposal was and wasn’t. This was not a government regulator. Hassabis was not proposing a statutory body with powers to ban AI models. The framework was designed as a pragmatic halfway house — industry-funded, technically expert, and capable of setting shared safety standards without waiting years for legislation to pass.

Demis Hassabis, Chief Scientist at Alphabet and co-founder of Google DeepMind, said, in the July governance essay that prompted the lobbying effort: “We need a Frontier AI Standards Body that can conduct rigorous technical evaluations of the most powerful AI systems before they are widely deployed.”

The 30-day pre-release evaluation window was specifically designed to be light enough not to block deployment, while creating a structured moment for independent review. Whether that balance would have held in practice is a separate question — but the proposal was not the heavy regulatory intervention its opponents implied.

The Broader Regulatory Picture

The stalling of this proposal doesn’t resolve the underlying question of how the US will handle AI oversight. The Trump administration is still actively debating its approach, and the Hassabis framework remains shelved rather than dead. Meanwhile, the UK Government has been developing its own path, including through the AI Safety Institute and ongoing work with international partners, though that effort is entirely separate from the US proposal.

Meanwhile, the global regulatory picture remains fragmented. The EU’s AI Act is moving through implementation. The UK has opted for a sector-by-sector approach rather than a single AI law. And the US, the country home to most of the world’s leading frontier AI labs, has yet to settle on any formal oversight model.

What This Means for Kent Residents

Kent residents and local public bodies — including Kent County Council and NHS Kent and Medway Integrated Care Board, both of which use data-driven and AI-supported tools in areas such as transport planning, social care and health analytics — rely on software and platforms developed largely by US companies. When the US fails to establish a formal safety standards framework for frontier AI, UK and EU regulators, along with bodies such as the Information Commissioner’s Office and the Competition and Markets Authority, become the primary source of assurance for the tools those organisations buy and deploy. Kent businesses in logistics, manufacturing and financial services using AI tools and APIs from American providers may need to lean more heavily on UK regulatory guidance and their own internal risk management, at least until the US position becomes clearer.