Favor slowing frontier development when other leading labs worldwide verifiably do the same; a temporary pause is an option within that coordinated approach. [1][2][3]
Initial statement. Favor slowing frontier development when other leading labs worldwide verifiably do the same; a temporary pause is an option within that coordinated approach.
Source [1] · Coauthored essay, updated edition
When AI builds itself
Attributed to: Jack Clark · Anthropic Institute
Published
“If such systems existed, we expect that we would slow down or temporarily pause”
What should we do?, paragraph beginning These systems would enable frontier AI developers · View original source ↗
Source context. Explicitly coauthored by Marina Favaro and Jack Clark. Current original edition visibly contains Update9/18/2026. September18 is verified edition/update publication, not asserted original utterance date. Essay positively favors effective slowdown, while offering temporary pause as a possible mechanism.
Scope
Frontier AI development at labs near the frontier across multiple countries.
Conditions
Global coordination tools must verify that other leading developers slow or stop under the same conditions and cannot secretly defect; unilateral slowdown may leave less cautious actors ahead.
Stated rationale
Explicitly coauthored by Marina Favaro and Jack Clark. Current original edition visibly contains Update9/18/2026. September18 is verified edition/update publication, not asserted original utterance date. Essay positively favors effective slowdown, while offering temporary pause as a possible mechanism.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 18 Sep 2026.
Accessed 3 Oct 2026.
Source [2] · Supporting original statement
What if automating AI R&D triggers an intelligence explosion?
Attributed to: Jack Clark · arXiv / authors
Published
“First, policymakers should develop ways to pace and constrain scale-ups of automated AI R&D.”
PDF page7, Steering and constraining an intelligence explosion, first numbered policy component · View original source ↗
Source context. Name appears on first-page author roster; paper explicitly says views are authors own, not employers. arXivv1 datesSeptember28. This is coauthored support for tools and managed capability growth, not a standalone demand for an immediate all-lab halt. Benefits and safety R&D still valued.
Scope
Scale-ups of automated AI R&D and the risk of software-driven intelligence explosion.
Conditions
Governments should consider safety/input conditions, capability growth limits per period, verification of agreements, incident responses and targeted workload pauses; weigh abuse and costs of delayed progress.
Stated rationale
Name appears on first-page author roster; paper explicitly says views are authors own, not employers. arXivv1 datesSeptember28. This is coauthored support for tools and managed capability growth, not a standalone demand for an immediate all-lab halt. Benefits and safety R&D still valued.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 28 Sep 2026.
Accessed 3 Oct 2026.
Source [3] · Supporting original statement
NPR Morning Edition: collective action problem
Attributed to: Jack Clark · NPR / KPBS
Published
“The issue we're talking about is there's a collective action problem here”
Quote immediately following discussion of voluntary Anthropic slowdowns · View original source ↗
Source context. Original NPR interview reporting by Steve Inskeep and Majd Al-Waheidi, hostedKPBS. Article datedSept15 andupdatedsame day. Recordingdateunknown. Supports coauthored essay context; export controls alone do not establish acceleration.
Scope
Collective industry development and international coordination, contrasted with action at one lab.
Conditions
Common safety standards, external scrutiny, cooperation between rivals; details of past company slowdowns undisclosed.
Stated rationale
Original NPR interview reporting by Steve Inskeep and Majd Al-Waheidi, hostedKPBS. Article datedSept15 andupdatedsame day. Recordingdateunknown. Supports coauthored essay context; export controls alone do not establish acceleration.
Source details
primary · Original publisher or author; verified independently of scout output.
Discovered 3 Oct 2026.
Published 15 Sep 2026.
Accessed 3 Oct 2026.
Editorial assessment. Research proposal only; independent adversarial review and explicit human approval required. Contrary/scope evidence: He advocates preserving a U.S. lead and has not disclosed details of prior Anthropic slowdowns. Geopolitical/export views do not negate conditional coordinated slowdown. Essay endorses slowdown option; not an unconditional immediate training halt. Remaining limitation: Exact original essay publication day not shown on rendered original; use clearly labeled September18 updated edition.