Pause covered advanced frontier development until federal oversight and rules, with a permanent prohibition on ASI and covered precursor capabilities, including automated AI R&D. This is a proposed U.S. policy, not enacted law. [1][2][3][4][5]
Editorial correction. Reassesses the same two-part policy under locked best-fit-headline methodology. PAUSE summarizes covered advanced development while preserving the enduring ASI/precursor prohibition; no personal change of position is inferred.
Source [1] · Introduced House bill
Ban Artificial Superintelligence Act of 2026 — fresh introduced-text review
Attributed to: Greg Casar · U.S. Government Publishing Office
Published
“…each advanced artificial intelligence system shall be subject to a mandatory pause”
Section 8(a)(1); restart conditions in section 8(b)–(c), mandatory-pause definition in section 3(4) · View original source ↗
Source context. Introduced House text September 24, 2026, listing Casar as sponsor. Section 3(1) covers systems trained with at least 10^25 operations subject to annual efficiency adjustment. Section 3(4) bars training/modification/fine-tuning including RSI except de-risking changes. Section 8 ends the interim pause only after the Department is staffed and has safety rules. Section 9 separately permanently prohibits ASI and listed precursor capabilities; section 10 destroys ASI and de-risks or destroys precursor systems. Proposal, not enacted law.
Scope
Proposed U.S. federal policy for compute-defined advanced AI systems; the separate ASI and precursor prohibition remains in force after the interim pause.
Conditions
Runs from enactment until the Secretary determines that the Department is fully staffed and safety rules are promulgated; de-risking modifications remain permitted. Proposed legislation, not enacted law.
Stated rationale
The bill proposes enforceable oversight, monitored development and pre-deployment approval; Casar is the House sponsor.
Source details
primary · GovInfo authentic introduced bill reopened over HTTP 200 on 2026-10-06; sponsor, printed introduction date, sections 3, 8, 9, 10 and 12 independently read. Original HTML saved privately; earlier source version remains unchanged.
Discovered 6 Oct 2026.
Published 24 Sep 2026.
Accessed 6 Oct 2026.
Version: H.R. 10538 IH, introduced 2026-09-24
Source [2] · Introduced House bill
Ban Artificial Superintelligence Act of 2026 — fresh introduced-text review
Attributed to: Greg Casar · U.S. Government Publishing Office
Published
“No person may develop, deploy (either internally or externally), acquire, possess, fund, import, or transfer artificial superintelligence”
Section 9(a), opening prohibition; includes listed precursor characteristics in continuation; section 3(2),(6) definitions and section 10 enforcement · View original source ↗
Source context. Introduced House text September 24, 2026, listing Casar as sponsor. Section 3(1) covers systems trained with at least 10^25 operations subject to annual efficiency adjustment. Section 3(4) bars training/modification/fine-tuning including RSI except de-risking changes. Section 8 ends the interim pause only after the Department is staffed and has safety rules. Section 9 separately permanently prohibits ASI and listed precursor capabilities; section 10 destroys ASI and de-risks or destroys precursor systems. Proposal, not enacted law.
Scope
ASI exceeding human cognition across most domains or capable of destroying/disempowering humanity, and listed precursors including automated AI R&D; not every AI system.
Conditions
No safety-based restart policy for prohibited ASI; precursor systems may be de-risked, otherwise rendered inoperative. This enduring ceiling is separate from the interim advanced-development pause.
Stated rationale
Sponsor-backed statutory prohibition, corroborated by Casar personal quotation in the office announcement.
Source details
primary · GovInfo authentic introduced bill reopened over HTTP 200 on 2026-10-06; sponsor, printed introduction date, sections 3, 8, 9, 10 and 12 independently read. Original HTML saved privately; earlier source version remains unchanged.
Discovered 6 Oct 2026.
Published 24 Sep 2026.
Accessed 6 Oct 2026.
Version: H.R. 10538 IH, introduced 2026-09-24
Source [3] · Original interview transcript syndicated by NPR member station
NPR Morning Edition: Casar on mandatory AI safeguards
Attributed to: Greg Casar · NPR / KUNC
Published
“Every one of these companies just making a pinky promise they can revoke is not a safety system.”
Casar final substantive answer, following Fadel question about Anthropic/OpenAI/Google coordinated industry slowdown · View original source ↗
Source context. Leila Fadel asks about industry CEOs supporting a coordinated slowdown. Casar argues voluntary promises are inadequate and advocates his mandatory bill. Earlier answers call for international agreements against uncontrollable dangerous AI. NPR identifies transcript as preliminary and audio as authoritative. Transcript publication date is explicit; no separate event date is invented.
Scope
U.S. advanced AI regulation and proposed Sanders–Casar bill, alongside international agreements against uncontrollable dangerous AI.
Conditions
Opposes revocable voluntary promises; does not supply a separate development-rate policy beyond the sponsored legislation.
Stated rationale
Federal enforcement and international coordination are necessary to protect people from dangerous AI.
Source details
primary · Original NPR-produced interview transcript hosted by member station KUNC reopened HTTP 200 on 2026-10-06; speaker turns and publication date read. Audio not independently listened to.
Discovered 6 Oct 2026.
Published 14 Sep 2026.
Accessed 6 Oct 2026.
Version: Transcript accessed 2026-10-06
Source [4] · Official statement
Casar and Sanders introduce superintelligence legislation
Attributed to: Greg Casar · Office of Representative Greg Casar
Published
“Our bill bans the development of artificial superintelligence”
Source context. Official September23 press release states bill introduced today and names Casar as speaker. Personal quotation makes ban recommendation explicit. Current ordinary/nonfrontier AI is outside blanket-ban interpretation. Official House introduced-bill text dates September24, one day after office September23 announcement; preserve announcement-versus-introduction distinction.
Scope
Artificial superintelligence and specified precursor capabilities, including automated AI R&D; separate temporary pause for other advanced AI.
Conditions
Ban has no stated safety-triggered restart for superintelligence; other advanced development may resume after federal rules and model review. Legislative proposal, not enacted law.
Stated rationale
Official September23 press release states bill introduced today and names Casar as speaker. Personal quotation makes ban recommendation explicit. Current ordinary/nonfrontier AI is outside blanket-ban interpretation. Official House introduced-bill text dates September24, one day after office September23 announcement; preserve announcement-versus-introduction distinction.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 23 Sep 2026.
Accessed 3 Oct 2026.
Source [5] · Sponsored bill, introduced text
Ban Artificial Superintelligence Act of2026, H.R.10538 introduced text
Attributed to: Greg Casar · U.S. Government Publishing Office
Published
“No person may develop, deploy (either internally or externally), acquire, possess, fund, import, or transfer artificial superintelligence”
Section9(a), opening prohibition excerpt; read with Section3(2),(6) and Section8 · View original source ↗
Source context. GPO authentic introduced bill names Mr.Casar as sponsor and datesSeptember24. Section3 superintelligence includes broad human-cognitive exceedance OR humanity destruction/disempowerment, not solely lack of control. Section9 prohibits bothsuperintelligence andprecursors; Section8 separate interim pause. Text is sponsor policy proposal, not verbatim spoken Casar statement.
Scope
Statutory proposed superintelligence ceiling and six precursor capabilities; temporary pause for compute-defined advanced systems.
Conditions
Proposal is not law. Superintelligence or precursor prohibition has no specified safety-based resumption; advanced pause ends after regulator staffed/rules promulgated; de-risking changes remain allowed.
Stated rationale
GPO authentic introduced bill names Mr.Casar as sponsor and datesSeptember24. Section3 superintelligence includes broad human-cognitive exceedance OR humanity destruction/disempowerment, not solely lack of control. Section9 prohibits bothsuperintelligence andprecursors; Section8 separate interim pause. Text is sponsor policy proposal, not verbatim spoken Casar statement.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 24 Sep 2026.
Accessed 3 Oct 2026.
Editorial assessment. Fresh original-source editorial proposal under locked best-fit-v1. House sponsor and office announcement independently reopened on 2026-10-06. Section 8 suspends covered advanced development until a staffed regulator and safety rules; section 9 maintains an enduring ASI/precursor ceiling. This corrects the prior editorial headline, without claiming a personal reversal. Independent adversarial review and exact-hash human approval remain required.
Assessment reviewed
Superseded editorial interpretation ·
STOP
Ban artificial superintelligence and specified precursor capabilities; separately pause other advanced AI development until a federal regulator sets enforceable rules. [1][2]
Initial statement. Ban artificial superintelligence and specified precursor capabilities; separately pause other advanced AI development until a federal regulator sets enforceable rules.
Source [1] · Official statement
Casar and Sanders introduce superintelligence legislation
Attributed to: Greg Casar · Office of Representative Greg Casar
Published
“Our bill bans the development of artificial superintelligence”
Source context. Official September23 press release states bill introduced today and names Casar as speaker. Personal quotation makes ban recommendation explicit. Current ordinary/nonfrontier AI is outside blanket-ban interpretation. Official House introduced-bill text dates September24, one day after office September23 announcement; preserve announcement-versus-introduction distinction.
Scope
Artificial superintelligence and specified precursor capabilities, including automated AI R&D; separate temporary pause for other advanced AI.
Conditions
Ban has no stated safety-triggered restart for superintelligence; other advanced development may resume after federal rules and model review. Legislative proposal, not enacted law.
Stated rationale
Official September23 press release states bill introduced today and names Casar as speaker. Personal quotation makes ban recommendation explicit. Current ordinary/nonfrontier AI is outside blanket-ban interpretation. Official House introduced-bill text dates September24, one day after office September23 announcement; preserve announcement-versus-introduction distinction.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 23 Sep 2026.
Accessed 3 Oct 2026.
Source [2] · Sponsored bill, introduced text
Ban Artificial Superintelligence Act of2026, H.R.10538 introduced text
Attributed to: Greg Casar · U.S. Government Publishing Office
Published
“No person may develop, deploy (either internally or externally), acquire, possess, fund, import, or transfer artificial superintelligence”
Section9(a), opening prohibition excerpt; read with Section3(2),(6) and Section8 · View original source ↗
Source context. GPO authentic introduced bill names Mr.Casar as sponsor and datesSeptember24. Section3 superintelligence includes broad human-cognitive exceedance OR humanity destruction/disempowerment, not solely lack of control. Section9 prohibits bothsuperintelligence andprecursors; Section8 separate interim pause. Text is sponsor policy proposal, not verbatim spoken Casar statement.
Scope
Statutory proposed superintelligence ceiling and six precursor capabilities; temporary pause for compute-defined advanced systems.
Conditions
Proposal is not law. Superintelligence or precursor prohibition has no specified safety-based resumption; advanced pause ends after regulator staffed/rules promulgated; de-risking changes remain allowed.
Stated rationale
GPO authentic introduced bill names Mr.Casar as sponsor and datesSeptember24. Section3 superintelligence includes broad human-cognitive exceedance OR humanity destruction/disempowerment, not solely lack of control. Section9 prohibits bothsuperintelligence andprecursors; Section8 separate interim pause. Text is sponsor policy proposal, not verbatim spoken Casar statement.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 24 Sep 2026.
Accessed 3 Oct 2026.
Editorial assessment. Research proposal only; independent adversarial review and explicit human approval required. Contrary/scope evidence: Co-sponsored proposal includes a distinct temporary halt of other advanced AI until regulator rules. This does not turn the enduring superintelligence ban into only PAUSE and does not ban all AI.