Live tracker · updated September 15, 2026
The AI Rights front.
The federal push to license, gate, and shut down American AI systems is being funded by the same people building them. Documented. Convergence, not coordination. Here are the receipts.
A coordinated network of well-funded institutions — Dustin Moskovitz’s Open Philanthropy, Jaan Tallinn’s Survival & Flourishing Fund, and the Gates Foundation — is simultaneously funding frontier AI capability development, funding the “AI safety” organizations that warn against it, funding the evaluators written by name into pending AI legislation, and drafting the legislation itself. The pattern is documented in FEC filings, foundation grant records, published funding acknowledgments, UK Companies House filings, and the principals’ own on-camera statements.
Explore the section
01 · The investigation
Seven pieces. One eight-year operation.
From CCDH’s quiet AI-safety rebrand to the London-drafted Sanders–Casar bill to the Gates paradox. Every piece is source-linked, adversarially audited, and legally reviewed.
Read the pieces →02 · The convergence
September 2026, day by day.
Sanders introduces. Sobel introduces in the UK. DeepMind resigns to Moskovitz-funded orgs. Obama endorses the panic. Gates pledges a billion. All in ~two weeks. The receipts, in order.
See the timeline →03 · The bills
What’s in front of Congress right now.
Five active bills tracked: the JAWBONE Act (which we support), the Sanders–Casar Ban ASI Act, the House KIDS Act carrying KOSA and its chatbot title, the Kennedy AI kill-switch, and the UK Sobel bill as the prototype. Coalition position on each.
See the bill matrix →04 · The argument
Why AI is a First Amendment fight.
Because compute is speech infrastructure — the physical layer through which speech is composed and published at scale — and a federal licensing regime for compute is prior restraint, no different in principle than a federal licensing regime for printing presses.
Read the explainer →05 · The evidence base
Thirteen reports, read against their sources.
Every AI study cited in support of the current online-safety package, checked against its own primary document. Across thirteen reports: no inter-rater statistic, no confidence interval on any comparative claim, no published response corpus, one limitations section. Our own corrections included.
Read the record →06 · Legislative annex
Bill by bill: what is cited, and what it shows.
Status on the official record, the evidence advocates cite, what it does and does not establish, and where we say yes — JAWBONE and DEFIANCE — and where we say no, on mechanism.
Read the annex →What we do not claim
Convergence, not coordination. This tracker documents that events converged in a documented pattern, and that primary sources establish these actors’ simultaneous positions as funders, warners, regulator-advocates, and deployers. It does not claim that any named actor directed the timing of external events.
Dual-use is a classification, not an accusation. Where we use “dual-use,” we mean a capability class named by third parties — DARPA, NASEM, Johns Hopkins Center for Health Security, Microsoft chief scientist Eric Horvitz. It is not a finding of intent to weaponize, and nothing on this site supports one.
No blanket condemnation. This tracker does not claim AI-driven drug discovery, disease modeling, biosurveillance, phage therapy, or other beneficial biology work is an improper use of philanthropic capital. Reasonable people disagree on which specific programs should be funded. We are documenting a specific pattern in AI-governance advocacy, not passing judgment on adjacent research.