Support slowing development of AI capabilities when their risks exceed society's ability to manage them.

Cory A. Booker · New Jersey · Democratic

policy impact 0.78 specificity 0.74 extraction confidence 88%

Commitment kind: action. Promised action: Support putting the brakes on AI capabilities whose risks exceed current risk-management capacity.. Promised outcome: Dangerous AI capabilities are slowed or paused until society can manage their risks.. Deadline: unknown; not assumed expired. As U.S. Senator from New Jersey during a 2026 Senate floor speech calling for a special AI session of Congress.

Occurrences

And yes, put the brakes on the development of capabilities whose risks outrun our ability to manage them

Booker said policymakers should slow or halt development of AI capabilities whose risks cannot yet be managed.

Booker Calls on Trump to Convene Congress for Special Session on AI
primary · speech · model gpt-5.5

Evidence

verified · Source version 63000 · locator 6657

And yes, put the brakes on the development of capabilities whose risks outrun our ability to manage them

Fetched verbatim passage. AI summary (separate from source): During the lookback window, Booker’s official Senate office published prepared floor remarks directly endorsing slowing AI capabilities when risks exceed society’s ability to manage them.

delivered same_term A for effort

Booker Calls on Trump to Convene Congress for Special Session on AI
primary · model gpt-5.5 · confidence 98%

unverified · Source version 87355 · locator unknown

Back then, I advocated for rigorous predeployment testing of advanced models, independent evaluations and audits

Unverified model/legacy excerpt; not proof. AI summary (separate from source): The Congressional Record documents Booker’s Senate floor statement describing prior AI safety positions, including predeployment testing and independent evaluations.

partial same_term A for effort

CREC-2026-09-17.pdf
secondary · model gpt-5.5 · confidence 88%

unverified · Source version 87377 · locator unknown

if the dangers of AI outpace our ability to manage them, then we need to slow down

Unverified model/legacy excerpt; not proof. AI summary (separate from source): In a lookback-window television interview, Booker again stated the commitment in near-identical terms, supporting slowdown when AI dangers outpace manageability.

delivered same_term

CNN.com - Transcripts
secondary · model gpt-5.5 · confidence 94%

verified · Source version 87379 · locator 1467

if the ability to control AI is outpaced by the threats, we should do things to slow it down

Fetched verbatim passage. AI summary (separate from source): A second lookback-window interview transcript shows Booker repeating support for slowing AI if risks outpace control, though he did not explicitly endorse an outright pause.

delivered same_term

Sen. Cory Booker: "Call Congress Back... We Need To Do Someting Bold" About AI | Video | RealClearPolitics
secondary · model gpt-5.5 · confidence 90%

unverified · Source version 87380 · locator unknown

which of AI’s emergent capabilities should be cultivated and which should be contained

Unverified model/legacy excerpt; not proof. AI summary (separate from source): Booker’s earlier official response to the White House AI strategy RFI shows a sustained position that some AI capabilities may need containment.

partial same_term A for effort

senator_cory_booker_response_to_wh_ostp_rfi_re_national_ai_strategy.pdf
secondary · model gpt-5.5 · confidence 87%

Assessments

Public state: delivered As of 2026-10-09 · verified · . Original AI recommendation (not independently established): delivered. same_term A for effort

The verified commitment was framed as supporting action to slow or brake AI capabilities when risks exceed society's ability to manage them, not as a promise that a binding federal pause would be enacted. Verified passages show Booker himself made that support explicit in official Senate remarks and reiterated that if control is outpaced by threats, AI should be slowed. Because the promised action was support, not enactment of a moratorium, these candidate-attributable statements satisfy the commitment in the same term. There is no verified evidence here of a broader implemented policy outcome, so the finding is limited to the support commitment.

provider codex_cli · model gpt-5.5 · confidence 90%