---
type: "post"
id: "h-r-10981-introduced-in-house-ih-3b52ee"
slug: "h-r-10981-introduced-in-house-ih-3b52ee"
title: "H.R. 10981 Introduced in House (IH)"
description: "Two House members introduced the American AI Security Act, a bill that would require federal rules for certain foundation models and pre-deployment review by the National Security Agency."
retrieval_nugget: "Two House members introduced the American AI Security Act, a bill that would require federal rules for certain foundation models and pre-deployment review by the National Security Agency. If enacted, it could make model access, safety-test records and compliance with federal recommendations conditions of deployment for models defined as covered."
published_at: "2026-11-16"
updated_at: "2026-11-16"
record_date: "2026-11-16"
date_kind: "published_at"
topics: ["ai-security"]
entities: ["American AI Security Act","H.R. 10981","Josh Gottheimer","Mike Lawler","National Security Agency","U.S. Department of Commerce","U.S. Department of Defense"]
related_patterns: []
novelty_label: "new_fact"
source_url: "https://www.govinfo.gov/content/pkg/BILLS-119hr10981ih/html/BILLS-119hr10981ih.htm"
source_urls: ["https://www.govinfo.gov/content/pkg/BILLS-119hr10981ih/html/BILLS-119hr10981ih.htm"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:h-r-10981-introduced-in-house-ih-3b52ee"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/h-r-10981-introduced-in-house-ih-3b52ee/","markdown":"https://newruntime.com/posts/h-r-10981-introduced-in-house-ih-3b52ee.md","json":"https://newruntime.com/posts/h-r-10981-introduced-in-house-ih-3b52ee.json"}
markdown_url: "https://newruntime.com/posts/h-r-10981-introduced-in-house-ih-3b52ee.md"
---

# H.R. 10981 Introduced in House (IH)

## Retrieval answer

Two House members introduced the American AI Security Act, a bill that would require federal rules for certain foundation models and pre-deployment review by the National Security Agency. If enacted, it could make model access, safety-test records and compliance with federal recommendations conditions of deployment for models defined as covered.

## Summary
Two House members introduced the American AI Security Act, a bill that would require federal rules for certain foundation models and pre-deployment review by the National Security Agency. If enacted, it could make model access, safety-test records and compliance with federal recommendations conditions of deployment for models defined as covered.

## What happened
Representatives Josh Gottheimer and Mike Lawler introduced H.R. 10981 in the House on November 10, 2026. The bill was referred to the Committee on Energy and Commerce and the Permanent Select Committee on Intelligence. Its short title is the American AI Security Act. It is a proposal, not an enacted law.

The bill directs the Secretary of Commerce, coordinating with the Secretary of Defense through the NSA Director and other named officials, to issue regulations no later than six months after enactment. Those regulations would define a “covered foundation model.” The criteria must include a minimum parameter threshold and capabilities that could pose a serious national-security risk, including a cybersecurity threat, or materially increase an individual’s, organization’s or foreign nation’s ability to develop a chemical, biological or radiological weapon. The bill defines a foundation model as a broadly trained, general-purpose software component adaptable to a wide range of tasks.

Before the initial deployment of a covered model, and before deployment after a substantial modification, its developer would have to give the NSA Director full access to the model and information about how it was trained, safety tests and their results, intended deployment, and intended recipients of access. The regulations would define what counts as a substantial modification. A developer could not deploy unless the NSA Director, in consultation with the relevant officials, had evaluated the model and made recommendations that the developer implemented, or the agency had failed to complete the required process within the specified period.

The NSA would have 30 days after receiving access and information to evaluate capabilities related to cybersecurity threats and chemical, biological or radiological weapons, and develop prevention recommendations. It could extend the evaluation by 30 days if it notified the developer in writing within the initial 30-day period. The NSA would then provide an unclassified report summarizing the evaluated capabilities and recommendations; it could include a classified annex. The bill exempts these evaluations from the Freedom of Information Act. The Secretary of Commerce could impose civil penalties of $500,000 per day for specified failures to provide access or violations of deployment regulations.

The bill would also establish an Article I Artificial Intelligence Security Review Court to hear developer appeals concerning NSA recommendations. Its review would be limited to whether the agency followed the bill’s procedures and kept recommendations within the evaluated capabilities; it could not review the methods, tools or techniques used in classified evaluations. The bill sets deadlines for appeals and briefs, and permits a Supreme Court review of certain decisions. It would preempt state or local requirements imposing specified transparency, disclosure or reporting obligations concerning covered models.

## What the archive adds
New Runtime’s earlier coverage described the AI Cyber Defense Act, introduced by Representatives Pat Harrigan and George Whitesides, as a proposal for NIST to develop voluntary standards addressing AI distillation attacks ([November 2, 2026 note](https://newruntime.com/posts/congressman-pat-harrigan-introduces-legislation-to-protect-american-97f5a1/); [November 4, 2026 note](https://newruntime.com/posts/harrigan-and-whitesides-introduce-ai-cyber-defense-act-86eed7/)). H.R. 10981 is a new bill, not a restatement: it proposes mandatory rules, NSA model access and deployment conditions for a category of models yet to be defined.

## Trend signal
This is a distinct legislative approach to frontier-model security, but it does not yet fit a pattern New Runtime tracks. The bill proposes government evaluation and controls on deployment; it does not establish a runtime boundary within an agent system or describe a shipped capability for builders to adopt.

## Why now
The proposal sets out a specific federal process for models that regulators would classify as covered, including evaluations before initial deployment and following substantial modification. If enacted, regulations would be due within six months, so the bill makes scope-setting and review timelines immediate questions for model developers and policymakers.

## Connections to other events
The proposal shares a broad concern with the AI Cyber Defense Act covered in New Runtime’s earlier notes: reducing national-security risks associated with advanced AI. The mechanisms differ. The earlier bill would direct NIST toward voluntary standards on distillation attacks; H.R. 10981 would create a regulatory process for covered-model evaluations, recommendations and deployment eligibility.

## Why it matters
For teams building or releasing foundation models, the proposal would make compliance a potential release gate rather than a post-launch reporting exercise. Developers of covered models would need to plan for government access to the model and detailed information about training, safety testing and intended access, then account for recommendations before deployment. The bill also reaches substantial modifications, so a release process could need a trigger for determining whether a change requires a new evaluation.

The practical impact depends on the definitions and regulations that would follow. A parameter threshold and capability criteria would determine which models are covered; the bill leaves those details to rulemaking. The proposed review deadlines, civil penalties and specialized appeal process indicate how the bill would structure oversight, but do not resolve how quickly an evaluation could be completed in practice or how developers would handle recommendations that affect a release.

## What happens next
- **The bill’s committee process will clarify whether Congress advances this model of mandatory pre-deployment review.** (6 months, 35%). Watch for: committee hearings, amendments or a reported bill.

## Limits and uncertainty
H.R. 10981 has only been introduced; its requirements are not in effect. The proposal leaves the covered-model threshold, capability criteria and definition of substantial modification to future regulations. It does not establish how many models would qualify, how much review would cost, or whether the 30-day evaluation period would be workable at scale. The text also does not establish what recommendations the NSA would make in a real case or how often developers would appeal. Those outcomes depend on enactment, subsequent rulemaking and implementation.

## Reader takeaway
The bill’s central proposal is a deployment gate: models defined as covered would face NSA evaluation and recommendations before launch or after substantial modification. For builders, the consequential details are not settled yet. Watch whether the proposal advances and how regulators would define coverage, because those choices determine whether the process affects a particular model or release.

## Open question
How would regulators define the covered-model threshold and substantial modification without sweeping in routine model updates?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Ai Security](https://newruntime.com/topics/ai-security.md): Explore the ai-security topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
