---
id: "2026-07-23-cats-act-ai-security-information-sharing"
url: "https://postcutoff.com/e/2026-07-23-cats-act-ai-security-information-sharing/"
as_of: "2026-10-23T13:24:00+02:00"
date: "2026-07-23"
date_precision: day
category: policy-safety
importance: 3
confidence: medium
status: [Partly confirmed]
sources: 2
editor: human editor
human_review: null
version: null
---

As of: 2026-10-23 13:24 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-07-23-cats-act-ai-security-information-sharing/

# CATS Act: AI labs may share threat data and coordinate

Full title: Bipartisan CATS Act introduced: antitrust safe harbour for AI labs to share threat information on distillation attacks and to coordinate on delaying or limiting deployment of high-risk models

In July 2026 Reps. George Whitesides (D-CA) and Bob Latta (R-OH) and Sens. Adam Schiff (D-CA) and Jim Banks (R-IN) introduced the Collaboration on Adversarial Threats and Security Risks (CATS) Act. Modelled on the Cybersecurity Information Sharing Act of 2015, it lets AI developers and other non-federal entities share information about AI security risks, including foreign "distillation attacks" on US models, and coordinate to reduce them, including by delaying or limiting deployment of high-risk models after written notice, with anti-competition guardrails. Anthropic listed the bill in its record Q3 2026 lobbying disclosure.

## Key facts
- Sponsors: Reps. George Whitesides (D-CA-27) and Bob Latta (R-OH-5); Sens. Adam Schiff (D-CA) and Jim Banks (R-IN). House cosponsors: Pat Harrigan, Darrell Issa, Ted Lieu, Mariannette Miller-Meeks, Nathaniel Moran, Jay Obernolte
- Date: the House press release is undated (bill text in a July 2026 folder); search results date Schiff's release and press coverage to July 23, 2026. Bill numbers not given on the page
- Provisions (Whitesides press release): statutory exemption for good-faith sharing of information on covered AI security risks, solely to prevent, investigate or mitigate them; coordination to reduce risks, 'including delaying or limiting deployment of high-risk AI models', only with prior written notice; coordination through bodies such as the NSA's AI Security Center, NIST's CAISI and the Frontier Model Forum; anti-competition guardrails and an Attorney General injunction power against abuse
- Endorsers: AI Policy Network, Alliance for a Better Future, Alliance for Secure AI, America First Policy Institute, ACT, CAIS Action Fund, Encode AI, FAI Action, Google, Health-ISAC, Law Reform Institute, SeedAI, SIIA
- Quotes: Schiff, 'We cannot allow China, or any foreign adversary, to steal American technological advancements in AI.' Encode AI's Sunny Gandhi: 'Developers need to be able to share threat information and coordinate, and this bill lets them do that.'
- Related bills found in the same search, not yet checked in detail: Deterring American AI Model Theft Act (H.R. 8283, Rep. Bill Huizenga, introduced April 15, 2026 per FDD Action) and the Senate BLADE Act (Blocking Large-Scale Adversarial Distillation Efforts, Sens. Andy Kim and Bill Hagerty, Aug 6, 2026)

## What happened

A bipartisan group in both chambers introduced the CATS Act to remove antitrust uncertainty for AI companies that share threat
intelligence. The trigger was distillation, in which foreign labs harvest outputs of US models to train their own, but the text also
covers joint decisions to hold back or limit the release of high-risk models.

## Why it matters

Labs have said antitrust law stops them from coordinating on safety, for example agreeing not to ship a dangerous capability. This bill
would create a narrow legal path for that, which is why it appears in the labs' lobbying in Q3 2026. It is backdated here: the dataset
first met it in an October lobbying disclosure (written partly from search summaries; the Senate release was not readable for us).

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 84 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 23 days after its cutoff
- Gemini 4 Argon (training cutoff March 2026): 114 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 53 days after its cutoff

## Sources

1. [Rep. Whitesides: Whitesides, Latta, Schiff and Banks introduce bipartisan bill to combat AI distillation and other attacks](https://whitesides.house.gov/?p=17548) (whitesides.house.gov, official)
2. [Sen. Schiff press release (July 23, 2026)](https://www.schiff.senate.gov/?p=13233) (schiff.senate.gov, official)

## Changes

- 2026-10-23 (filed): Created

## Related

- 2026-10-21: [Record Q3 lobbying by Nvidia, Anthropic and OpenAI](https://postcutoff.com/e/2026-10-21-ai-labs-record-q3-lobbying-antitrust-information-sharing/index.md)
- 2026-10-12: [Anthropic says new defenses cut the Claude reasoning harvested by distillers by about 90 to 99 percent](https://postcutoff.com/e/2026-10-12-anthropic-distillation-defenses/index.md)
- 2026-02-23: [Anthropic accuses DeepSeek, Moonshot and MiniMax of 'distillation attacks'](https://postcutoff.com/e/2026-02-23-anthropic-distillation-attacks-report/index.md)
- People: [Ted Lieu](https://postcutoff.com/person/ted-lieu/)