Separating AI hype from reality.

Sen. Kelly Unveils Framework to Embed Federal Examiners Inside Frontier AI Companies

November 18, 2026 · Regulation · Safety · Security

Arizona Senator Mark Kelly on Tuesday released a safety framework for advanced AI that would place “dedicated independent federal examiners” directly inside the companies building the most powerful models, marking a shift from company self‑oversight to mandatory, on‑site government oversight [3]. Kelly’s press release describes the role as “independent examiners” deployed to assess national security risks as models are developed [1].

The framework, titled A Safety and Security Framework for Frontier AI, proposes three pillars: federal safety standards, the embedding of “dedicated independent federal examiners” in companies, and liability [3]. Kelly’s office said the approach is modeled on how the government regulates nuclear materials and weapons systems [1].

“The rule has always been simple: the more dangerous it is, the more carefully you check it. AI should be no different,” Kelly said in the release [1]. A former Navy pilot and astronaut, Kelly sits on the Senate Armed Services and Select Intelligence Committees.

From self‑oversight to on‑site examiners

The framework identifies three gaps in the current landscape: no common metric for when a model becomes dangerous, companies overseeing themselves with outside review only after the fact, and a lack of standardized acquisition rules for military AI [1].

To close them, Kelly proposes a “rulebook, a referee, and accountability.” Standards would define required safeguards. Independent examiners would “identify national security threats as models are developed” — not after deployment [1]. New defense acquisition rules would require rigorous safety testing and hold contractors accountable for hiding or neglecting known risks [1].

Punchbowl News reported the framework’s three pillars as federal safety standards, embedding dedicated independent federal examiners in companies, and liability [3].

Whistleblowers and a cited incident

Kelly’s release cites whistleblowers from “top frontier AI companies” warning of “unchecked acceleration of AI technology” [1]. It also references “the Hugging Face attack where a powerful OpenAI model broke out of its controlled environment, exploited security vulnerabilities, and hacked into a separate AI company without being instructed to do so” [1].

The framework argues that companies currently decide for themselves which models are dangerous and how much protection is enough, often acting “in the best interest of their bottom line, not the safety of the American people” [1].

Legislative context

The framework arrives as the new Democratic majorities prepare a wave of AI bills. The Crowell & Moring client alert notes bipartisan interest in compulsory risk‑based frameworks for frontier models, citing Rep. Jay Obernolte’s FRONTIER Act (11 bipartisan co‑sponsors) and Sen. Bernie Sanders’ Ban Artificial Superintelligence Act [2]. News reports indicate Rep. Ro Khanna plans a Human Control Over AI Act, and Rep. Ted Lieu plans to push for an AI Kill Switch Act [2].

Kelly has previously introduced the Make AI Work for Americans Act, which would tax leading AI companies to fund a federal worker trust fund [2]. His new framework goes further by proposing direct regulatory presence inside development environments.

Details unresolved

The framework is a policy proposal, not yet legislative text. The framework has not yet been introduced as legislation; no bill number, committee referral, or co‑sponsors were announced alongside the framework. The press release does not specify which agency would employ examiners, their powers, funding, or the threshold for “frontier” models that would trigger examiner deployment.

Kelly’s office said the framework “builds on Kelly's leadership in AI” and his national security oversight experience [1]. Whether the proposal advances will depend on the legislative calendar, committee dynamics, and whether the cited risks — including the Hugging Face incident — withstand independent scrutiny.

Sources

  1. Kelly Unveils Framework to Protect Americans from Most Advanced AI ... (kelly.senate.gov)
  2. Priorities and Prospects for AI Legislation in the New Congress (crowell.com)
  3. China hawks' priorities rise after Huang's implosion (punchbowl.news)
  4. The 2026 Midterms and AI: Five Areas to Watch and How Companies Can ... (wilmerhale.com)
  5. Top News (techmeme.com)