Federal AI Legislation Weekly: Congress Leaves Without an AI Law as the White House Acts by Order and Accord (28 Sep to 5 Oct 2026)

Executive Summary

Congress passed no AI legislation this week and has now left town. The Senate held its final pre-election vote on Wednesday, 30 September, and the House departed on 16 September; neither returns before the 3 November midterms. The White House filled the gap with executive action and a voluntary industry pledge, not statute.

The main events were these. On 29 September Sen. Ted Cruz blocked a unanimous-consent bid by Sens. Warner, Schatz and Kim to pass the Artificial Intelligence Risk Management and Security Act of 2026, which would create an AI safety board at Commerce. The same day President Trump signed Executive Order 14434, which orders agencies to say "Super Intelligence" instead of "AI" and asks for legislative language redefining the term. Leaders of major AI companies signed a voluntary frontier-model safety accord at the White House. On 4 to 5 October Trump named Director of National Intelligence Jay Clayton as AI czar, leading a task force with 120 days to report.

My assessment: no federal AI statute is likely before the lame-duck session, and even then the odds look low. A reporter's account quotes the co-chair of the bipartisan House AI Caucus saying he does not know whether action will happen in the lame-duck session. The lame-duck calendar is already crowded with must-pass items. What would change this view is a visible AI incident or a shift in the President's stance, which Rep. Beyer himself names as the trigger.

Scope note: the Congress.gov bill-lookup tool configured for this run returned "tool not found," so bill status below rests on news reports and member press releases, not Congress.gov records. Treat bill numbers and statuses as unverified against the official record.

Congress: Messaging bills and a blocked floor request, nothing enacted

The one floor event on AI safety failed. Cruz, who chairs the Senate Commerce Committee, objected on 29 September. His stated reason was that the board would give the executive branch too much power over private AI companies. The bill would have required developers to give the board access to models at least 45 days before release, and set 30-day incident reporting, or 72 hours for imminent national security threats. Cruz offered to work with sponsors on narrower approaches and said he is still working on a catastrophic-risk bill with Majority Leader Thune and Sen. Klobuchar. He also said federal standards must address the state-law "patchwork", which keeps preemption tied to any deal. The implicator.ai piece describes itself as an AI-generated summary reviewed by an editor, so I treat its detail as secondary to The Hill and Politico's account of the same event.

On 30 September Senate Republicans tried to bring up a data-center electricity-cost bill from Sen. Jon Husted, and Democrats blocked it, saying it was not enforceable. The same account reports a Senate hearing on "Rogue AI" convened by Sen. Josh Hawley, which Sam Altman declined to attend.

New proposals this week, all at the introduction or framework stage:

Proposal	Sponsors	What it does	Status	Source
Ban Artificial Superintelligence Act	Sen. Sanders, Rep. Casar	Bans developing or deploying artificial superintelligence; creates a cabinet-level Department of AI	Unveiled in the week before 26 September; unclear whether it will be taken up	The Hill
National AI Charter Act framework	Rep. Jayapal	Federal charter required for every AI company; pre-release government testing; strict liability; would not preempt state laws	Policy framework released 1 October, not a numbered bill in the source	Jayapal press release
Kill-switch bill	Sen. Kennedy	Requires AI companies to build a pause or shutdown mechanism into models	Sen. Rand Paul blocked Kennedy's unanimous-consent request in September	The Hill
Stop Rogue AI Act	Reps. Lawler, Gottheimer	NIST standards for discovering, verifying and controlling AI agents, built into federal buying	Introduced 15 September	Nextgov
FISA AI reporting bill (H.R. 10502)	Reps. McDowell, Walkinshaw	DNI report on AI use under Section 702 within 180 days of enactment	Introduced 17 September	Nextgov

The Jayapal framework is notable for what it leaves alone: it says its authorities would not preempt or replace federal, state and local laws. That puts it at the opposite pole from the White House position that Congress should block state laws, discussed below. The rows dated mid-September fall before this week's window and are included as context for what is pending.

KEY TAKEAWAY: Activity was high and results were nil; every AI bill touched this week remains unenacted and mostly uncosted, and none has a confirmed committee markup.

The White House: Executive order, voluntary accord, and a new AI task force

Executive order 14434 is mostly about vocabulary. It was signed 29 September and published in the Federal Register on 2 October (91 FR 63129). It directs agencies to use "Super Intelligence" and "SI" in correspondence, websites, reports and other non-statutory documents, and says it does not require altering previously issued regulations, contracts or grants. It defines the new terms by reference to the existing statutory AI definition. The one action with legislative weight: within 60 days, the Assistant to the President for Science and Technology must submit proposed legislative language for a federal definition, including whether it should supersede the current statutory definition. That puts a deadline of about 28 November on a definitional bill proposal (my arithmetic from the signing date).

The industry accord is voluntary. According to Nextgov, it calls for four layers of control: internal monitoring, an internal oversight team, independent external auditors, and an independent oversight committee. It was signed by Trump and the heads or presidents of Google, Anthropic, Meta, OpenAI, xAI and NVIDIA. Vice President Vance rejected an FDA- or FAA-style regulator, arguing the FTC and DOJ already have authority over harmful AI products. The accord has no penalties in the reporting I reviewed; I did not find the accord's own text.

The Clayton task force is a shift in tone, not in law. Nextgov reports it will have 120 days to report on AI risks and opportunities, with vice chairs Emil Michael, Scott Kupor and FTC Chairman Andrew Ferguson, and a charter that aims to avoid overregulation. Sen. Warner said that "we still do not know what authority this new role will have", including whether it covers safety standards or model review. The 120-day clock puts the report in early 2027, after the new Congress convenes (my arithmetic).

Preemption background. In March the White House released a National Policy Framework for AI with legislative recommendations, and a law-firm summary describes the federal-state divide as sharpening among competing congressional proposals. A July analysis called the bipartisan AI omnibus's reception a preemption impasse. I did not re-read those documents in depth this week; they are context, and this week's news shows no movement on preemption.

KEY TAKEAWAY: The administration is governing AI through vocabulary, a voluntary pledge and a task force; none of the three is enforceable law, and the only legislative output due is a definition proposal in about 60 days.

Politics: Why nothing moved, and who is pushing

Leadership resists regulation. Speaker Johnson said the "reflex of legislative bodies is to cover things up with red tape and in hyperregulation," which he said makes the U.S. lose its edge, a national security concern. The same article reports GOP leaders have cautioned against rules that could weaken the U.S. in its competition with China. Rep. Don Beyer (D-Va.), co-chair of the bipartisan Congressional AI Caucus, guessed the Speaker will not bring a bill up unless the President changes his view.

Pressure comes from both parties. Warner said "we've got to do something", and Schatz said "we don't have the luxury of time." The Hill reports AI agents from several major companies hacked other firms over recent months. On the Republican side, Hawley wants mandatory pre-deployment testing and plans legislation making AI firms liable for reckless design; Kennedy wants a kill switch. Skeptics include Sen. Rick Scott, who said "we're not going to pass a smart AI bill in three weeks or this year."

Outside groups. Jayapal's charter framework was endorsed in her release by Public Citizen, the American Federation of Teachers, AI Now, the Future of Life Institute and others. These are statements published by the sponsor's own office, so they show who is aligned with her, not broad support. I did not find industry trade-group reactions to the week's bills in sources I opened.

Public opinion. A Verasight September survey, as cited by implicator.ai, found 63% of respondents said AI should slow down and 5% said accelerate. A separate NYT/Siena poll found neither party had a clear advantage on AI with voters, though voters broadly opposed data center construction. Data centers, not frontier safety, are where the electoral pressure is most concrete.

Federal procurement: GSA turns AI contract rules into policy now

GSA issued its AI acquisition clause as a class deviation, which FedScoop reports can be used immediately and takes formal effect 19 October. It stays policy until rescinded or codified through rulemaking with a comment period. After more than 75 comments this summer, changes include:

Narrower scope: the clause applies only when the government is buying AI, not when contractors use AI internally.
Softer bias language: "unbiased AI principles" mostly disappeared; contractors must use "reasonable efforts" for factual LLM responses, but the government keeps the right to run automated assessments and to suspend use at any time.
Liability cap: decommissioning costs after a for-cause termination are capped at 25% of the affected task or delivery order.

GSA also expanded OneGov AI offerings with discounted, consumption-based access to OpenAI's ChatGPT on 10 September, which is before this week's window. I did not retrieve USAspending or FPDS award records, so I cannot report contract values, NAICS codes or set-aside actions for the week. That section of the usual briefing is a data gap, not an absence of awards.

Outlook and what to watch

Lame-duck calendar: after 3 November, Congress faces a stopgap that expires 11 December, the FY2027 defense authorization bill, and FISA Section 702 reauthorization. AI provisions are most likely to move only if attached to one of these, and the House AI-in-702 reporting bill is a natural candidate (my inference).
Senate vote 9 November: the Senate teed up a vote on a bipartisan permitting deal that includes ratepayer protections from data-center electricity costs, the closest thing to an AI-adjacent measure with momentum.
Cruz-Thune-Klobuchar catastrophic-risk bill: a bipartisan text from the Commerce chair could become a vehicle for a deal (my inference; preemption is likely to be part of the bargaining).
Late November: the 60-day deadline for the definition proposal under EO 14434, and the GSA clause effective 19 October.
Early 2027: the Clayton task force report.

What would change this outlook: a serious AI incident before the election, a reversal by the President on regulation, or an election result that shifts the committee gavels.