Senate’s AI Select Committee Proposal Owns the Investigation, Not the Legislation

S.Res. 918 was submitted and referred to committee on September 29, 2026. It has not been voted on, heard, or marked up. What the text reveals about its own design is worth reading carefully.

S.Res. 918 has been submitted and referred to committee. There has been no vote, no hearing and no markup, and the Select Committee described here does not exist. The Artificial Intelligence Safety Board bill in the same Record, S. 5576, was introduced by Senators Warner, Schatz and Kim. It is not Senator Cruz’s bill. Nothing here is investment advice.

What Happened

On September 29, 2026, Senator Gallego submitted S.Res. 918 to the Senate floor. The resolution was referred to the Committee on Rules and Administration. There has been no vote, no hearing, and no markup. Most resolutions referred to committee do not advance further.

The resolution would create a Senate Select Committee on Artificial Intelligence, composed of not more than thirteen Senators appointed by the President pro tempore. It would sit during the 120th Congress — meaning nothing in it takes effect in the session in which it was introduced.

The same day’s Congressional Record, page S5173, also carries S. 5576 — a separate bill to establish an Artificial Intelligence Safety Board. That bill was introduced by Senator Warner, joined by Senators Schatz and Kim, and referred to the Committee on Commerce, Science, and Transportation. It is not Senator Cruz’s bill. Senator Cruz appears in the day’s proceedings separately, reserving the right to object during floor consideration of distinct AI risk-management business. The Record keeps those two things apart, and so does this piece.

The Structural Read

Jurisdiction is the binding constraint on any new congressional committee. Commerce, Judiciary, Armed Services, Banking, and Homeland Security already hold pieces of artificial intelligence between them.

A committee that claimed exclusive referral rights would have to take that jurisdiction from colleagues who currently hold it. That is the fight that kills most proposals of this kind.

Section 4(b)(2) removes the fight by giving up the prize. Here is the verbatim language from the resolution:

The trade is legible. Maximum investigative reach. Minimum legislative control. The committee can compel a witness to appear. It cannot compel a bill to come to it.

Whether that is the right trade is not for this piece to say. It is a description of how the resolution is drafted.

The Definition Question

Section 1 of the resolution defines “artificial intelligence” by reference. The term has the meaning given in section 5002 of the National Artificial Intelligence Initiative Act of 2020, codified at 15 U.S.C. 9401. That is the same statutory section that the Super Intelligence executive order issued on the same day said it was not changing.

Both instruments rest on a definition written in 2020.

The resolution then does something the executive order did not. It writes a second definition from scratch — this one for “artificial intelligence system.” That term means any machine-based system that, for explicit or implicit objectives, infers from the input it receives how to generate outputs, including predictions, content, recommendations, or decisions, that can influence physical or virtual environments.

The resolution inherits the older term and drafts a newer one beside it. The two do not cover the same ground. That is a drafting choice with consequences.

The Bottom Line

S.Res. 918 is a resolution submitted in one Congress, designed for the next, drafted to avoid the one fight it could not win — and equipped with subpoena power precisely because it gave up everything else. That is not a criticism. It is a description of the only architecture that can clear a field already carved into pieces by five standing committees. Whether the 120th Congress finds the design useful is a question for a political environment that has not yet been elected.