Washington's AI Rulebook Is a Pledge, a Probe, and a Task Force With No Clear Job

On September 29, executives from OpenAI, Anthropic, Meta, Google, Nvidia, and xAI stood in the White House East Room and signed what President Trump called "almost a constitution" for artificial intelligence. Ten days later, reporting on the anniversary of that signing shows a federal posture that looks less like a constitution and more like three overlapping, uncoordinated experiments: a voluntary pledge with no penalties, an open Federal Trade Commission investigation into two of the pledge's signatories, and a newly announced "Super Intelligence Force" whose statutory authority — if it has any — remains undefined. None of the three has been reconciled with the others, and that gap is quickly becoming the real substance of U.S. AI policy.

A pledge with no penalties, and no clear path to one

The pact itself is thin by design. It asks the six signatories to maintain internal safety controls, stand up an internal oversight team, retain an outside auditor, and create a board-level committee to review results — but it requires no public disclosure of what those audits find, gives federal agencies no enforcement role, and leaves implementation entirely to the companies. Trump has described it as "morally binding" and pointed to what he called "tremendous self-policing" in the industry. The document itself hedges on its own durability, noting it may eventually "make sense to codify these steps into laws or regulations" — an admission that the administration is treating voluntary compliance as a placeholder, not an end state.

Notably, the split over that placeholder runs through the industry itself, not just outside critics. OpenAI's head of global affairs, Chris Lehane, has called for the pact to be a floor rather than a ceiling, arguing industry standards should "complement — not replace — mandatory federal safeguards and democratic oversight." Anthropic CEO Dario Amodei has gone further, publicly calling for "more serious and binding regulation of AI." That two of the six signatories are on record wanting more enforcement than the document they just signed provides is itself a signal to boards and investors: the companies most exposed to frontier-model liability do not think self-attestation will hold up, legally or politically, for long.

The same week, a federal investigation into two signatories

The contradiction sharpens around timing. Within roughly 24 hours of the pact signing, the FTC opened a broad inquiry into whether OpenAI and Anthropic — both pact signatories — violated federal unfair-and-deceptive-practices law, with plans to issue formal information demands and seek executive testimony. FTC Chair Andrew Ferguson, appointed by the same administration that hosted the pact ceremony, framed the inquiry around accountability for AI tool behavior rather than novel AI-specific rules, saying he avoids "anthropomorphizing of these tools" and that liability should attach to what a tool actually does. The result is an administration simultaneously treating AI safety as something companies should self-certify and something the government is actively investigating two of those same companies over — using 1970s-era consumer-protection authority because no AI-specific statute exists to use instead.

All that they're getting is grace from this government. — Sarah Myers West, AI Now Institute, on the voluntary pact

Enter the Super Intelligence Force

Layered on top of both is the Super Intelligence Force, announced by Trump via executive order and Truth Social in late September and detailed through early October. It is chaired by Director of National Intelligence Jay Clayton — also floated as a potential AI czar, though the administration has not confirmed the title — alongside FTC Chair Ferguson, Pentagon research-and-engineering official Emil Michael, and Office of Personnel Management Director Scott Kupor. Its stated job is to "coordinate the Federal Government's engagement" with consumers, public interest groups, religious organizations, critical-infrastructure operators, and AI companies to keep the U.S. ahead of China. What it does not have, based on the order's own text and the administration's public statements, is enumerated regulatory power, a budget, or a defined relationship to either the voluntary pact or the FTC's active investigation. Reporting on the order has noted it arrives bundled with an effort to rename AI itself as "super intelligence," a rebrand critics read as narrative management rather than governance.

Whether the Force becomes the oversight body Trump has separately floated — he mentioned after the pact signing that he was considering a committee to "watch over the whole entity" — is, by the administration's own account, still an open question. A body that might end up supervising the pact's self-reported audits is chaired in part by the same FTC official currently investigating two of the pact's signatories under separate authority. If that overlap gets formalized, it would concentrate soft-power oversight, enforcement discretion, and industry-relationship management in the same small set of hands without a statute defining how those roles interact.

What this means for compliance teams and investors

For companies building or deploying frontier models, the practical reality is that there is currently no single federal AI rulebook to comply with — there are three parallel tracks, each run on different legal footing, each capable of changing faster than legislation could. The FTC's unfair-and-deceptive-practices authority is real and already being used; a formal 6(b)-style information demand carries genuine legal exposure regardless of what any voluntary pact says. The pact itself carries reputational and competitive risk — a company that under-invests in the audits it voluntarily committed to risks becoming the next FTC target, pact or no pact. And the Super Intelligence Force, while toothless today, sits one executive order away from becoming an operational chokepoint if its mandate is clarified toward the oversight role Trump has floated.

Investors should also track the legislative undercurrent this patchwork is meant to preempt. A Sanders-Casar bill would impose criminal penalties on individuals for reckless superintelligence development; a separate bipartisan AI Systems Transparency Act would mandate disclosure of training data and safeguards regardless of what any company pledges voluntarily. Both remain long shots in the current Congress, but their existence is the backdrop against which the administration's voluntary, discretionary approach is being pitched — and the pressure valve that determines how long "self-policing" can substitute for statute before Congress, courts, or a major AI-caused incident forces the issue.