Prohibits superintelligence and self-improving precursors; linked policy requires scientific safety consensus and public buy-in before lifting the ban. Other AI and data centres remain allowed. [1][2][3][4]
Clarification. Dated evidence assessment; subsequent source clarifies scope and conditions without inferring reversal.
Source [1] · Authored essay
America’s Security Depends on Stopping Superintelligence
Attributed to: Connor Leahy · Connor Leahy / ControlAI
Published
“It should prohibit the development of superintelligence within the US, including immediate precursors like autonomously self-improving AIs”
Source context. Personal July1 essay recommends global prohibition and explicitly excludes banning all AI development. Both prohibition links point to the conditional original superintelligence statement. Lack of repeated restart wording in the essay does not erase incorporated conditions.
Scope
Development of superintelligent AI and immediate precursors such as autonomously self-improving AI, domestically and globally.
Conditions
The authored July1 essay twice links its recommended prohibition to the superintelligence statement: not lifted before broad scientific consensus on safe controllable development and strong public buy-in. No fixed restart date; author regards superintelligence inherently uncontrollable, but does not explicitly reject those linked conditions.
Stated rationale
Personal July1 essay recommends global prohibition and explicitly excludes banning all AI development. Both prohibition links point to the conditional original superintelligence statement. Lack of repeated restart wording in the essay does not erase incorporated conditions.
Source details
primary · Original publication date verified; separate utterance date unknown and not substituted.
Discovered 4 Oct 2026.
Published 1 Jul 2026.
Accessed 5 Oct 2026.
Source [2] · Official testimony
Canadian Senate testimony on AI security
Attributed to: Connor Leahy · Senate of Canada
Statement
“It should be illegal to attempt to build superintelligence.”
Answer to Senator Aucoin after the opening statement; paragraph immediately following arrest and pause language. · View original source ↗
Source context. Senate identifies Leahy as U.S. Director of ControlAI and former Conjecture CEO. April 15 hearing includes pause wording but also explicit illegality and worldwide commitments not to build. It separately preserves current AI use and rejects slowing Western data centres.
Scope
Development of extremely powerful general-purpose superintelligence.
Conditions
International trust-but-verify prohibition enforced domestically; no restart condition stated in this answer; not a ban on all AI or data centres.
Stated rationale
Senate identifies Leahy as U.S. Director of ControlAI and former Conjecture CEO. April 15 hearing includes pause wording but also explicit illegality and worldwide commitments not to build. It separately preserves current AI use and rejects slowing Western data centres.
Source details
primary · Official transcript heading and bilingual PDF establish hearing April 15, 2026 at 6:48 PM ET. Transcript publication day not supplied; do not substitute April 17 social upload.
Discovered 4 Oct 2026.
Accessed 4 Oct 2026.
Source [3] · Official testimony
Canadian Senate testimony — data-centre distinction
Attributed to: Connor Leahy · Senate of Canada
Statement
“…the slowdown of the development of data centres”
Answer to Senator Arnold about disengagement and Matt Shumer, after discussion of international buy-in. · View original source ↗
Source context. This is a deliberate contrary-scope check in the same hearing. Infrastructure opposition is not a personal CONTINUE recommendation for frontier capability scaling.
Scope
AI use and data-centre development in Western countries, distinct from superintelligence training.
Conditions
Reject disengagement that shifts power to rogue jurisdictions; his superintelligence prohibition remains.
Stated rationale
This is a deliberate contrary-scope check in the same hearing. Infrastructure opposition is not a personal CONTINUE recommendation for frontier capability scaling.
Source details
primary · Official transcript hearing April 15, 2026. Separate publication day unknown.
Discovered 4 Oct 2026.
Accessed 4 Oct 2026.
Source [4] · Authored essay
A Roadmap for Cognitive Software and A Humanist Future of AI
Attributed to: Connor Leahy · Connor Leahy and Gabe Alfour / Conjecture
Published
“We can build a future where cognitive software is understandable, reliable, and truly beneficial to humanity.”
Source context. Coauthored Connor Leahy and Gabe Alfour, original December 2, 2024. Product/research roadmap supports useful AI but cannot establish acceleration or contradict scoped prohibition.
Scope
Reliable cognitive software and engineered AI tools.
Conditions
Humanist software-engineering approach rather than uncontrolled black-box replacement systems; no frontier tempo stated.
Stated rationale
Coauthored Connor Leahy and Gabe Alfour, original December 2, 2024. Product/research roadmap supports useful AI but cannot establish acceleration or contradict scoped prohibition.
Source details
primary · Original publication date verified; separate utterance date unknown and not substituted.
Discovered 4 Oct 2026.
Published 2 Dec 2024.
Accessed 4 Oct 2026.
Editorial assessment. Private revision after independent rejection of STOP: original July essay twice links conditional prohibition. Preserve strong global ban language and inherent-uncontrollability belief, but do not infer permanent no-restart from omission of restated conditions. No new personal stance reversal claimed.
Assessment reviewed
Earlier position ·
STOP
Prohibits superintelligence through enforceable international commitments; not all AI or data centres. [1][2][3]
Initial statement. Initial private assessment based on original dated evidence; not public editorial approval.
Source [1] · Official testimony
Canadian Senate testimony on AI security
Attributed to: Connor Leahy · Senate of Canada
Statement
“It should be illegal to attempt to build superintelligence.”
Answer to Senator Aucoin after the opening statement; paragraph immediately following arrest and pause language. · View original source ↗
Source context. Senate identifies Leahy as U.S. Director of ControlAI and former Conjecture CEO. April 15 hearing includes pause wording but also explicit illegality and worldwide commitments not to build. It separately preserves current AI use and rejects slowing Western data centres.
Scope
Development of extremely powerful general-purpose superintelligence.
Conditions
International trust-but-verify prohibition enforced domestically; no restart condition stated in this answer; not a ban on all AI or data centres.
Stated rationale
Senate identifies Leahy as U.S. Director of ControlAI and former Conjecture CEO. April 15 hearing includes pause wording but also explicit illegality and worldwide commitments not to build. It separately preserves current AI use and rejects slowing Western data centres.
Source details
primary · Official transcript heading and bilingual PDF establish hearing April 15, 2026 at 6:48 PM ET. Transcript publication day not supplied; do not substitute April 17 social upload.
Discovered 4 Oct 2026.
Accessed 4 Oct 2026.
Source [2] · Official testimony
Canadian Senate testimony — data-centre distinction
Attributed to: Connor Leahy · Senate of Canada
Statement
“…the slowdown of the development of data centres”
Answer to Senator Arnold about disengagement and Matt Shumer, after discussion of international buy-in. · View original source ↗
Source context. This is a deliberate contrary-scope check in the same hearing. Infrastructure opposition is not a personal CONTINUE recommendation for frontier capability scaling.
Scope
AI use and data-centre development in Western countries, distinct from superintelligence training.
Conditions
Reject disengagement that shifts power to rogue jurisdictions; his superintelligence prohibition remains.
Stated rationale
This is a deliberate contrary-scope check in the same hearing. Infrastructure opposition is not a personal CONTINUE recommendation for frontier capability scaling.
Source details
primary · Official transcript hearing April 15, 2026. Separate publication day unknown.
Discovered 4 Oct 2026.
Accessed 4 Oct 2026.
Source [3] · Authored essay
A Roadmap for Cognitive Software and A Humanist Future of AI
Attributed to: Connor Leahy · Connor Leahy and Gabe Alfour / Conjecture
Published
“We can build a future where cognitive software is understandable, reliable, and truly beneficial to humanity.”
Source context. Coauthored Connor Leahy and Gabe Alfour, original December 2, 2024. Product/research roadmap supports useful AI but cannot establish acceleration or contradict scoped prohibition.
Scope
Reliable cognitive software and engineered AI tools.
Conditions
Humanist software-engineering approach rather than uncontrolled black-box replacement systems; no frontier tempo stated.
Stated rationale
Coauthored Connor Leahy and Gabe Alfour, original December 2, 2024. Product/research roadmap supports useful AI but cannot establish acceleration or contradict scoped prohibition.
Source details
primary · Original publication date verified; separate utterance date unknown and not substituted.
Discovered 4 Oct 2026.
Published 2 Dec 2024.
Accessed 4 Oct 2026.
Editorial assessment. Private roster candidate. Original evidence supports a scoped recommendation; preserve contrary evidence and explicit conditions. No approval, portrait or public integration. Category is provisional pending independent adversarial review.