Anthropic tells Newsom courts should be able to block risky AI models
In an October 15 letter to California Governor Gavin Newsom, Anthropic proposed that officials be able to seek court orders blocking or suspending AI models posing serious risk, while opposing broad "kill switches". OpenAI, in a separate letter, urged mandatory shutdown controls instead.
1 / 2
The story, neutrally told
Centre · 1Anthropic told California Governor Gavin Newsom in an October 15 letter that government officials should be able to seek court orders blocking the release of AI models that pose a "serious risk" to the public. CNAN “government officials should be able to seek court orders blocking the release of AI models that pose a "serious risk" to the public” Read at CNA ↗ Centre · 1The letter said that if the attorney general concludes, informed by an independent evaluation, that a frontier company's models pose a near-term risk of catastrophic harm, the AG should be able to seek a court order to suspend a model's deployment, internal use or further development. ComputerworldN “they should be able to seek a court order to suspend a model’s deployment, internal use, or further development,” Anthropic wrote in its October 15 submission.” Read at Computerworld ↗ Centre · 1An Anthropic spokesperson confirmed the letter is authentic, and Newsom's office did not immediately respond to a request for comment; Politico first reported its details. CNAN “An Anthropic spokesperson confirmed the letter is authentic. The California governor's office did not immediately respond to a request for comment.” Read at CNA ↗
Centre · 1The submissions followed Newsom's September 18 executive order N-9-26, which directs the Government Operations Agency, with the Office of Emergency Services and national experts, to submit recommendations by November 16 on amending AI safety laws; the order does not itself impose the proposed requirements. ComputerworldN “directs California’s Government Operations Agency, in consultation with the Governor’s Office of Emergency Services and national experts, to submit recommendations by November 16” Read at Computerworld ↗ Centre · 2Anthropic said it opposes requirements to build "kill switches" that shut down entire models, and prefers targeted steps such as disabling a problematic AI agent or cutting off a specific customer's access; it called a "circuit breaker" a more apt metaphor. CNAN “it opposes requirements that developers build "kill switches" capable of shutting down entire AI models” Read at CNA ↗ ComputerworldN ““We think a more apt metaphor for what frontier AI companies need to build is a circuit breaker,” it wrote” Read at Computerworld ↗ Centre · 1Anthropic also proposed independent verification, including risk reports at least every six months, unredacted access for evaluators, and a state-administered pooled fund paid into by developers to finance evaluators. ComputerworldN “It proposed requiring developers to publish risk reports at least every six months”“developers pay into a state-administered pooled fund to finance evaluators, rather than pay them directly” Read at Computerworld ↗
Centre · 1In an October 20 letter, OpenAI took a different route, supporting an amendment to California's SB 53 to require frontier developers to maintain effective shutdown controls and an enforceable duty to use them; it also favoured the term "shutdown controls" over a single switch and recommended annual independent compliance audits. ComputerworldN “We support amending SB 53 to require frontier developers to maintain effective shutdown controls and an enforceable duty to use them when unsafe conditions arise”“OpenAI recommended annual independent compliance audits” Read at Computerworld ↗ Centre · 1Both companies back independent safety assessments and broader reporting of incidents in which models circumvent safeguards or take unauthorized actions. ComputerworldN “Both companies also called for broader reporting of incidents involving models that circumvent safeguards or take unauthorized actions.” Read at Computerworld ↗ Centre · 1CNA noted Anthropic has itself faced restrictions: in June the Trump administration ordered it to restrict foreign access to its cyber-capable Mythos 5 and Fable 5 models, and the Commerce Department later lifted the order. CNAN “The Trump administration in June ordered Anthropic to immediately restrict foreign access to its cyber-capable Mythos 5 and Fable 5 models” Read at CNA ↗
Centre · 1At the federal level, a bipartisan Senate group debated in September a bill letting the government block unsafe model releases, subject to challenge in federal court, but it has not advanced; House lawmakers proposed an "AI Kill Switch Act" in July. Anthropic is also preparing for a US IPO. CNAN “Congress left for its pre-election break shortly after, and lawmakers have not moved forward on that proposal.”“Anthropic is preparing for a US initial public offering.” Read at CNA ↗
Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.
Left0 outlets
No left outlet in our sources has covered this story yet.
Centre2 outlets
- Framing
- Both outlets report the submissions as policy proposals. CNA leads on Anthropic's letter with federal context and Anthropic's own past restrictions; Computerworld frames it as an Anthropic–OpenAI divergence on who should shut down risky AI, with an enterprise-advice angle.
- Emphasis
- CNA: court-order power, opposition to kill switches, federal bills, IPO. Computerworld: side-by-side comparison, independent verification, an analyst's advice to enterprises.
- Leaves out or plays down
- CNA does not mention OpenAI's competing proposal or the executive order's details; Computerworld omits the federal bills, Anthropic's June restrictions and the IPO.
- Charged language
- “kill switches”“pull the plug”
- For example
-
“Anthropic itself has faced government restrictions on its AI systems.” — CNA
“they are proposing sharply different approaches to stopping models that pose serious risks” — Computerworld
Right0 outlets
No right outlet in our sources has covered this story yet.
What every side reports
- Anthropic sent Newsom an October 15 letter proposing court power to block or suspend risky models.
- Anthropic favours targeted interventions over whole-model kill switches.
- Newsom's September actions prompted the submissions.
Where accounts differ
-
Who should have power to stop a risky model
- Centre
- Anthropic: officials/AG seeking a court order. OpenAI: developers legally required to maintain and use shutdown controls (per Computerworld).
Anthropic organisation
Wants courts empowered, on an attorney general's request, to suspend dangerous models; supports independent verification; opposes whole-model kill switches in favour of targeted "circuit breaker" controls.
“When a review finds a model poses a serious risk, the government should be able to ask a court to block it” — CNA
“We strongly support requiring independent verification of what frontier AI companies report” — Computerworld
OpenAI organisation
Argues developers should be required to maintain and use effective shutdown controls, via an SB 53 amendment, rather than rely on court orders.
“OpenAI argues developers should be required to maintain and use effective shutdown controls” — Computerworld
Gavin Newsom person
Convened an expert panel in September and issued an executive order seeking safeguards recommendations; his office had not commented on the letter.
“Anthropic submitted the recommendations after the governor in September convened a panel of artificial intelligence experts” — CNA
East and Southeast Asia1 outlet
CNA runs the Anthropic letter story with US federal and political context.
“Congress left for its pre-election break shortly after, and lawmakers have not moved forward on that proposal.” — CNA
CNA
North America1 outlet
Computerworld treats it as a comparison of AI labs' proposals with enterprise implications.
“For enterprises, Gogia said, contractual shutdown responsibilities should match the controls each party actually possesses.” — Computerworld
Computerworld
Left0 articles
No coverage yet.
Centre2 articles
-
CNAN ·
Anthropic tells California governor courts should be able to block risky AI models
Neutral News report on Anthropic's letter, adding federal legislative context and Anthropic's own past government restrictions.
-
Anthropic and OpenAI diverge on who should pull the plug on risky AI
Neutral Comparative explainer of Anthropic and OpenAI submissions with an analyst's enterprise guidance.
Right0 articles
No coverage yet.
- 28 Oct 16:08 First CNAN Anthropic tells California governor courts should be able to block risky AI models
- 29 Oct 12:12 +20h 4m ComputerworldN Anthropic and OpenAI diverge on who should pull the plug on risky AI
Times are when each article was published, or when we first saw it if the outlet gave no time.
Ask about this story
Loading…