Sponsor: Sen. Ted Cruz (R-TX) · Cosponsors: Sen. Ron Wyden (D-OR)
Status: Reported favorably by Senate Commerce, 18–10, on September 16, 2026, as amended by the Cruz substitute. Cantwell #1 — a substitute removing AI models as covered entities and striking the §§ 4–5 transparency portal — failed 13–15. Awaiting Senate floor scheduling; no date. Cruz committed to work with Cantwell pre-floor on harmonizing with a separate catastrophic-risk AI bill. Congress.gov had not yet recorded the committee action as of this update. Full record and next steps →
Creates a federal cause of action against government coercion of platforms, broadcasters, and AI systems — including attempted coercion, and with real discovery to uncover the record.
- Federal private right of action against federal officials who coerce speech intermediaries
- Reaches attempted coercion, not just successful takedowns
- Real discovery: uncover the informal-pressure record the courts have so far refused to reach
- Ongoing transparency requirements
Coalition analysisThis is the good bill in the room. It reaches the exact informal federal pressure that dismissed
Finn v. GEC on procedural grounds and that the current OSTP FOIA is refusing to disclose under (b)(5). It is viewpoint-neutral. It is bipartisan. Sayer Ji has
publicly endorsed it as founder.
Sponsor: Sen. Bernie Sanders (I-VT) · Cosponsors: Rep. Greg Casar (D-TX) — House companion
Status: Recently introduced. Bill tracker link pending Congress.gov assignment. Text architecture matches the UK Sobel Bill and the UK Online Safety Act.
Ban on “artificial superintelligence” with corporate death penalty shutdown authority and an independent evaluator framework naming METR-class organizations.
- Federal authority to shut down American AI developers deemed to have crossed a capability threshold
- Independent evaluator framework (METR is Moskovitz-funded, and Moskovitz sits on the METR board)
- Extraterritorial reach implied
- Architecture drafted by ControlAI — funded by Tallinn, who is also an Anthropic investor
Coalition analysisA federal statute that lets the government shut down American AI developers is prior restraint on the modern printing press. Compute is speech infrastructure. The bill’s bipartisan framing is a political mechanism; the constitutional problem is the same regardless of sponsor. The drafters, the funders, and the beneficiaries are the same institutional pipeline documented in Pieces 2, 4, 5.
Sponsor: Rep. Brett Guthrie (R-KY-2), Energy and Commerce chairman · Cosponsors: 1 (Rep. Frank Pallone, D-NJ-6, ranking member, June 23, 2026)
Status: Passed the House 267–117 under suspension on June 29, 2026 (Roll no. 228). Received in the Senate July 13, 2026 and referred to Commerce, where the Senate’s own Kids Online Safety Act (S. 1748, Blackburn) was advanced by voice vote on August 5, 2026 alongside the CHATBOT Act (S. 4407, Cruz–Schatz). On September 30, 2026, Sen. Rand Paul objected to a unanimous-consent request to pass S. 1748, so there has been no Senate floor vote; the chambers have not reconciled. Full KOSA entry on the Policy Blueprint tracker.
A House omnibus, assembled from more than a dozen bills, that rewrites the Kids Online Safety Act without the Senate’s duty of care and adds the first federal design-and-disclosure mandate for AI chatbots, enforceable by the FTC and all fifty states.
- Chatbot providers must tell any user they “know or should have known” is a minor that the system is an AI, deliver suicide and crisis hotline information after related prompts, and prompt breaks after extended sessions
- Providers must adopt unspecified policies and procedures to “address” listed harms (sexual exploitation, gambling, promotion of drugs, tobacco, alcohol) — a standard that reaches lawful expressive uses such as satire, role-play and medical information
- Title II: platforms must “establish, implement, maintain, and enforce” policies against listed harms to minors and ship default safeguards for suspected minors; the “should have known” standard drives platform-wide age estimation
- Enforcement by the FTC and every state attorney general; preemption of stronger state laws was a Democratic objection at the March 5, 2026 markup (28–24)
Coalition analysisThe chatbot title is the AI Rights problem in miniature: government-prescribed design, government-compelled messages, and a vague harm standard with fifty enforcers — the statutory version of the informal pressure this coalition sued over. Unlike the Sanders–Casar bill it does not license compute, so it is not a prior restraint on the press itself; it is a mandate on how speech tools may be built and what they must say. The Senate companion path matters more: S. 1748 keeps a “duty of care” that
CCDH endorsed in 2024 and
still asks Congress for, and
X Corp negotiated the December 2024 text with the sponsors. We oppose the KIDS Act as passed and S. 1748 as reported; we would revisit a version limited to actual-knowledge, conduct-based rules with no state-AG speech enforcement.
Sponsor: Sen. Kennedy · Cosponsors: n.a.
Status: Introduction pending. Bill text not yet public. Coalition position will be posted here once text is available and analyzed against the same drafting-ecosystem lens applied to Sanders–Casar.
A federal “kill switch” authority over AI systems, described in reporting but with no public bill text yet.
- Kill-switch authority (scope TBD until text drops)
- Drafting connections TBD — we will apply the same due-diligence lens used on Sanders–Casar
Coalition analysisA federal kill-switch over AI systems raises the same prior-restraint concerns as the Sanders–Casar Ban ASI Act. We are withholding a full position until text is public. If the drafting ecosystem traces to the same funders and evaluators documented in Piece 4, we will oppose.
Sponsor: Feryal Clark MP (Labour) · Cosponsors: Labour AI policy leadership
Status: Introduced. Currently the working prototype for the US Sanders–Casar bill.
UK statute imposing independent-evaluator, shutdown-authority, and extraterritorial-scope regulation on AI providers — the same architecture as the UK Online Safety Act.
- Independent evaluator language (ControlAI-adjacent)
- Extraterritorial scope explicit
- Shutdown authority
- Drafted by the same ecosystem as CCDH/ControlAI — principal funder Jaan Tallinn (also an Anthropic investor)
Coalition analysisThis is the prototype. When it passes in the UK, the US bill is next. The transatlantic pipeline documented in Piece 5 means the UK bill is not a foreign story — it is the road map for what is being drafted for Congress.