On May 29, 2026, OMB published a proposed rule rewriting 2 CFR Part 200, the Uniform Guidance that governs every federal grant. The largest revision since 2013 includes a clause that would let any agency cancel any grant at any time if it decides the work is no longer in the “national interest.” For university AI labs running multi-year projects on NSF or NIH awards, the practical consequence is blunt: federal grant money becomes revocable working capital, not committed funding.
What the proposed rule actually does
The proposed rule would add an at-will termination clause to every grant agreement. Agencies would include language warning recipients that their award could be canceled at any time if the funding agency decides the grant is no longer in the national interest, according to Ars Technica’s reporting. No cause requirement. No appeal standard. The agency’s discretion is the entire basis.
Peer review would be explicitly subordinated under the proposal. According to Ars Technica, political appointees would have final say over grant decisions and were specifically instructed not to “routinely defer” to peer reviewers. Grant decisions would shift from expert program staff to political appointees, and every grant program would need to be aligned with administration priorities (Ars Technica).
The proposed rule would also restrict federal funding for several categories: disparate-impact theories, DEI-related activities, and certain foreign collaborations (NACo). Ars Technica describes the overall effect as bans on “culture war topics” and limits on international collaborations (Ars Technica).
Even the mechanics of doing research get tighter. Ars Technica reports the rule would block spending on publishing papers and attending conferences funded through grants (Ars Technica).
Why formal rulemaking, and why now
OMB tried this agenda once before, as an executive order in August 2025. Courts struck down the resulting grant cancellations because executive orders cannot circumvent statutory requirements. The administration lost multiple cases (Ars Technica). This time, OMB is using the formal rulemaking process under the Administrative Procedure Act, with a public comment period and Federal Register publication. The goal is to give these changes the force of regulation rather than executive fiat.
The structural mechanism matters. The proposed rule reclassifies 2 CFR Subtitle A from guidance to regulation (NACo). Future OMB amendments to these rules would take effect government-wide without requiring each agency to run its own separate rulemaking. One edit to Subtitle A, and every agency’s grant terms change automatically.
The rule also introduces new procedural machinery: pre-award review, integration with the Treasury’s Do Not Pay system for pre-payment verification, and a 10-day requirement for Inspectors General to disclose findings to a US Attorney (NACo). These are enforcement mechanisms, not just policy statements.
What at-will cancellation does to research lab economics
The second-order effect is not about any single grant. It is about how labs plan around federal money.
A typical university AI lab running an NSF or NIH award operates on a three- to five-year project plan. That plan commits to hiring PhD students and postdocs, reserving GPU compute clusters months ahead, and signing subcontracts with other universities or industry partners. These are long-horizon, hard-to-reverse commitments made against a funding schedule that the lab assumed was, if not permanent, at least governed by defined termination procedures.
At-will cancellation removes that assumption. If a grant can be pulled because an agency decides the work no longer aligns with current administration priorities, the lab director faces a risk that has no mitigation under existing institutional structures. There is no grant-cancellation insurance. The risk stays on the lab’s balance sheet.
The parallel to at-will employment is direct. When US labor law moved to at-will contracts, employers gained flexibility and workers absorbed the risk of arbitrary termination. The same transfer happens here: the agency gains discretion, and the grantee absorbs the cost of any cancellation that occurs after the lab has already committed people and compute to the project.
For AI research, the timing problem compounds. Compute allocations at shared research infrastructure are awarded on competitive cycles measured in months. A postdoc hired against a three-year grant who loses funding in year two loses the project, the trained researcher, and the compute allocation simultaneously. The sunk cost is mostly unrecoverable.
The foreign collaboration restrictions
The proposed limits on certain foreign collaborations hit AI research at a structural weakness. The field runs on international teams. Large model training, benchmark development, and safety research all involve cross-border collaborations that move faster than the grant cycle.
For labs that have built multi-year international collaborations around specific research agendas, the new restrictions create immediate uncertainty about whether those partnerships can continue under federal funding. The scope of which collaborations would be affected remains unclear from available reporting.
What to do during the comment period
The proposed rule is open for public comment via the Federal Register. For lab directors and university research offices, the actionable items are narrow:
Submit comments. The APA requires agencies to respond to significant comments. Comments that identify specific technical problems, such as the inconsistency between multi-year grant commitments and at-will termination, carry more weight than general opposition.
Audit current grant portfolios for exposure. Identify which active awards could be affected by the political-alignment requirement or the international collaboration restrictions. Labs with ongoing foreign partnerships on NSF or NIH grants should assume those collaborations will be reviewed.
Diversify funding stacks. If the proposed rule takes effect, the effective risk of any single federal grant increases. Labs that supplement with industry contracts, foundation grants, or institutional bridge funding will absorb cancellations better than labs running on a single federal award.
Push institutional leadership to engage. University government-relations offices exist for this kind of rulemaking. The comment period is open now, and coordinated institutional responses are still feasible.
The rule is proposed, not final. It will face public comment, likely legal challenges, and the standard friction of implementation. But the direction is clear: federal grant funding would be restructured from a commitment mechanism into a discretionary one. Labs that treat their current grant schedules as guaranteed multi-year funding are making a bet that the proposed rule, or something close to it, will not take effect. That is a bet with no hedge.
Frequently Asked Questions
Does the proposed rule affect currently active grants, or only new awards?
The proposed rule does not explicitly address whether termination clauses would be retrofitted into existing grant agreements. Because the rule reclassifies 2 CFR Subtitle A from guidance to regulation, its provisions would apply government-wide upon taking effect, but the gap on retroactivity remains an open question for the comment period (docket OMB-2026-0034, closing July 13, 2026). Labs with multi-year awards already in force should raise this ambiguity in comments.
How would the publication and conference cost restrictions work day to day?
Under current Uniform Guidance, publication costs are a standard allowable expense. The proposed rule would make them unallowable unless expressly required by statute or approved in advance by the Federal agency on a case-by-case basis, with the same restriction applied to conference travel. Each paper submission or conference trip would need individual agency sign-off before the expense is incurred, creating a per-item approval gate that does not exist today.
Which specific countries are named in the collaboration restrictions?
The proposed rule includes an outright ban on funding collaborations involving Chinese researchers and establishes a domestic-first framework for any international collaboration. This goes further than the general limits on foreign partnerships described in the article by naming China specifically and imposing a preference ordering that requires domestic researchers to be prioritized in any federally funded collaborative work.
What exact language does the rule use to demote peer review?
The rule states that peer review remains advisory and does not replace agency discretion. This formulation does not merely deprioritize expert input. It establishes that agency funding decisions cannot be challenged on the grounds that they contradict peer reviewer recommendations, creating a legal shield for political overrides that is stronger than a simple preference for alignment would be on its own.