Skip to main content
VerifiedLab commitment

Anthropic's When AI builds itself: the option to slow or pause, conditional on verifiable peers

Anthropic published When AI builds itself on 4 June 2026. It states that it would be good for the world to have the option to slow or temporarily pause frontier AI development so that societal structures and alignment research can keep up, and that Anthropic would slow or pause if other developers at or near the frontier did so in a verifiable manner. It is a conditional commitment by one company, binding nobody, and its condition has not been met.

The verdict

Verified

The document exists. The ledger fetched it at its publisher and quotes it.

Key facts

What the sources say

Record ID
DKT-2026-0017
Kind
Lab commitment
Jurisdiction
United States
Last verified
Added
  • Anthropic's text: it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up.
  • The commitment is conditional: Anthropic would slow or pause if other developers at or near the frontier also did so in a verifiable manner.
  • Anthropic names what a pause would need: multiple well resourced labs in multiple countries agreeing, each able to verify the others, plus specified triggers, lifting conditions and an adjudicator.
  • Scientific American reported on 5 June 2026 that the post dated the report to 4 June and that Anthropic said recursive self improvement could come sooner than most institutions are prepared for.
  • Senator Sanders' letter of 10 August 2026 quoted the companies' own past words about pausing back to them (DKT-2026-0002).

Dimension by dimension

7 dimensions, each one stated, silent or open

Binding, Stage, What it would require, What it does not do, Who is for it, Who is against it, Next step on the record. Stated means the document you can open below says it; silent means the ledger read the document and it does not.

BindingStated
No: a conditional statement of intent by one company.Anthropic, When AI builds itself, primary source, 4 June 2026.
StageStated
Published 4 June 2026; the condition, verifiable peer action, has not occurred as of 16 September 2026.Anthropic, When AI builds itself, primary source, 4 June 2026.
What it would requireStated
Of Anthropic, a slowdown or pause only if peers do the same verifiably; of anyone else, nothing.Anthropic, When AI builds itself, primary source, 4 June 2026.
What it does not doStated
It commits to no unilateral pause, sets no threshold and names no verifier.Anthropic, When AI builds itself, primary source, 4 June 2026.
Who is for itStated
Anthropic.Anthropic, When AI builds itself, primary source, 4 June 2026.
Who is against itStated
None named at publication; the ministry answer of 14 September 2026 (DKT-2026-0011) is the first governmental rejection of the coordination premise.Anthropic, When AI builds itself, primary source, 4 June 2026.
Next step on the recordStated
Amodei's essay of 12 September 2026 turned the conditional into a three step plan and a unilateral first step (DKT-2026-0005).Anthropic, When AI builds itself, primary source, 4 June 2026.

What it changes

For a policy or compliance lead

This is the document the later letters quote. A conditional pause commitment reads as a promise to legislators and as a hedge to critics, and both readings are on the docket. For a compliance lead the useful part is Anthropic's own list of what a pause needs, agreement, verification, triggers, an end condition and an adjudicator, because it is the same list you would need to evaluate any vendor's claim that it would stop if asked.

Sources

What this record was verified against

  1. Anthropic, When AI builds itselfPrimary · 4 June 2026
  2. Scientific American, Anthropic warns AI may soon begin recursive self-improvementSecondary · 5 June 2026

Related

Records that sit beside this one

Cite this record

Free to reuse under CC BY 4.0, with attribution. The record ID DKT-2026-0017 is permanent and is never reused.

In a sentence
According to the GAGE Slowdown Docket (as of 16 September 2026), anthropic's when ai builds itself: the option to slow or pause, conditional on verifiable peers.
APA
GAGE (Global Academy of Generative-AI Education). (2026). Anthropic's When AI builds itself: the option to slow or pause, conditional on verifiable peers. Slowdown Docket. Retrieved 16 September 2026, from https://www.gage.academy/tools/slowdown-docket/records/DKT-2026-0017-anthropic-when-ai-builds-itself-pause-option
MLA
"Anthropic's When AI builds itself: the option to slow or pause, conditional on verifiable peers." Slowdown Docket, GAGE (Global Academy of Generative-AI Education), 16 September 2026, https://www.gage.academy/tools/slowdown-docket/records/DKT-2026-0017-anthropic-when-ai-builds-itself-pause-option.
Chicago
GAGE (Global Academy of Generative-AI Education). "Anthropic's When AI builds itself: the option to slow or pause, conditional on verifiable peers." Slowdown Docket. Last modified 16 September 2026. https://www.gage.academy/tools/slowdown-docket/records/DKT-2026-0017-anthropic-when-ai-builds-itself-pause-option.
Permalink
https://www.gage.academy/tools/slowdown-docket/records/DKT-2026-0017-anthropic-when-ai-builds-itself-pause-option

Last updated . Every record re verified . The ledger is checked weekly, every Monday, and the same day for any bill, probe, hearing, letter or lab commitment.

Back to the full ledger, or every record for United States and every lab commitment record.

GAGE briefings tell you which AI regulation deadlines are coming, what they actually require of you, and when a program opens.

Powered by GAGE, free EU AI Act tools from the Global Academy of Generative-AI Education
CallEmailHelp