Coordinate a slowdown among frontier AI developers, including China; an initial proposal asks OpenAI and Anthropic to defer immediate recursive self-improvement for the next year. [1][2]
Initial statement. Coordinate a slowdown among frontier AI developers, including China; an initial proposal asks OpenAI and Anthropic to defer immediate recursive self-improvement for the next year.
Source [1] · Original broadcast transcript
This Week September13 2026 transcript: Jacob Coxon interview
Attributed to: Jacob Coxon · ABC News
Published
“…needing some sort of slowdown between the people building the different A.I.s”
Coxon segment; first answer to George Stephanopoulos asking about Amodei slowing down, paragraph beginning Yeah · View original source ↗
Source context. Original ABC rush transcript of September13 Sunday broadcast. Coxon says he definitely agrees with Amodei slowing down, then calls slowdown proposals sensible first steps. When asked about banning superintelligent AI, calls a ban something to discuss, not an unequivocal demand. Publication and broadcast September13; ABC warns rush transcript may be updated.
Scope
Coordinated frontier AI development pace among leading developers, including China.
Conditions
International coordination; slower development is a first step, with an initial two-lab RSI deferral retained as a concrete example.
Stated rationale
Original ABC rush transcript of September13 Sunday broadcast. Coxon says he definitely agrees with Amodei slowing down, then calls slowdown proposals sensible first steps. When asked about banning superintelligent AI, calls a ban something to discuss, not an unequivocal demand. Publication and broadcast September13; ABC warns rush transcript may be updated.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 13 Sep 2026.
Accessed 3 Oct 2026.
Source [2] · Original edited interview
Jacob Coxon: Crunch time for humanity
Attributed to: Jacob Coxon · WIRED
Published
“…some sort of neutral understanding that they won't immediately go into recursive self-improvement in the next year.”
Answer to What do you think needs to be done now?, second sentence (excerpt after WIRED editorial bracket) · View original source ↗
Source context. Original Maxwell Zeff interview explicitly transcribed questions and Coxon answers; edited for clarity/length. First sentence calls agreement a baby step. Immediately following answer discusses China and international pacing. PublicationSeptember9; recording date separately unstated.
Scope
Immediate recursive self-improvement at OpenAI and Anthropic; ultimately international pacing.
Conditions
Initial two-lab agreement; ultimately international coordination including China and understanding compute distribution. Not a halt of every AI activity or permanent rejection of benefits.
Stated rationale
Original Maxwell Zeff interview explicitly transcribed questions and Coxon answers; edited for clarity/length. First sentence calls agreement a baby step. Immediately following answer discusses China and international pacing. PublicationSeptember9; recording date separately unstated.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 9 Sep 2026.
Accessed 3 Oct 2026.
Editorial assessment. Research proposal only; independent adversarial review and explicit human approval required. Contrary/scope evidence: September13 ABC answer endorses coordinated slowdown as the first step; considers a ban an open discussion rather than a settled demand. September9 two-lab temporary deferral remains a concrete narrower measure within wider international pacing. Remaining limitation: ABC is an original rush transcript that may contain transcription errors; WIRED recording date unstated. No permanent all-AI ban established.