Politico reported on 27 October 2026 that Anthropic has told California Governor Gavin Newsom that courts should have the authority to block AI programs that pose serious risks, and that AI companies should develop kill switches and have outside safety auditors embedded alongside employees.
Politico says those positions were spelled out in a letter the company sent earlier this month to Newsom. We have not read the letter itself; the quotes below are the ones Politico reports. It is a set of proposals, not a law or a bill.
Short explainer · AI policy brief
A vertical news Short narrated from the published headlines and opening lines of this piece and four related FourWeekMBA AI policy pieces.
Business Pill · Least privilege
An evergreen Business Engineer explainer on least privilege: giving a system only the access its task needs. It teaches the general idea only and does not discuss Anthropic, OpenAI or California.
The key insight: As we read it, Politico’s account shows the two labs agreeing on embedded auditors and shutdown tools, and splitting on whether a government should be able to go to court to stop a model.
Politico reports that Anthropic reiterated its support for having third-party safety evaluators working inside frontier AI companies. Those evaluators “should be able to report risks they identify to the state Attorney General,” the letter said, according to Politico.
If the attorney general concludes that a company’s models pose a “near-term risk of catastrophic harm,” Politico reports, the attorney general should be able to seek a court order to suspend a model’s deployment, internal use, or further development.
On access, Politico reports that Anthropic compared what it grants evaluators today with what banks give outside auditors, saying “it should be comparable to what a company’s own risk assessment staff have.”
Who Pays the Auditors
What each lab’s letter to Gov. Newsom backs, per Politico: Anthropic’s three proposals (a court route to suspend a model, kill switches and embedded outside auditors) against OpenAI’s two (shutdown controls and embedded auditors), plus Anthropic’s six-month risk-report cadence. Count is ours. Source: Politico, 27 Oct 2026.
Politico reports that companies should pay into “a fund administered by the state” rather than paying the outside groups directly. State officials would rate the auditors on criteria like “the rigor of their reasoning” and their track record of identifying risks, Politico says.
That way, Politico reports, state authorities could require AI companies to use at least one highly rated evaluator, or assign them by random selection or rotation. Politico also reports that the letter says large developers should publish risk reports at least every six months, and that a law focused only on shutdown “could provide a false sense of reassurance.”
OpenAI Sent Its Own Letter
Politico reports that OpenAI followed with a similar letter to Newsom last week. In it, OpenAI detailed its support for “shutdown controls” on powerful AI models and for embedding independent safety auditors inside frontier AI companies, Politico says.
According to Politico, OpenAI did not go as far as Anthropic’s proposal to let governments block high-risk models.
What Newsom Asked For
Politico says the letters respond to ideas Newsom raised in an executive order last month, including “kill switch” shutdown procedures and independent safety monitors inside AI companies.
The Benton Institute’s summary of the 18 September order says it aims to “dramatically accelerate the implementation of new third-party oversight of safety and security risks in AI systems, as well as independent audits.”
The summary says a group of experts is to produce “within two months, a guide for California to reinforce and strengthen its AI safety and security laws.”
Politico reports that California’s Legislature is expected to pursue a slew of regulations when it returns in about a month.
Connecting the Coverage
On 17 October we reported Anthropic’s draft bill limits AI model blocks to $1B risks: a federal draft, reported by Politico, under which Commerce could ask a federal court in Washington, D.C. to suspend a frontier model only for a foreseeable, imminent risk of more than $1 billion in damage or more than 50 people killed or seriously injured. The Newsom letter, as Politico describes it, sets out a state-level route with a different trigger: the attorney general and a “near-term risk of catastrophic harm.”
As we read it, the letter moves the stop decision from the company to a judge: Politico reports the attorney general would seek a court order, on a “near-term risk of catastrophic harm.”
As we read it, the payment design targets the auditor’s incentive: Politico reports companies would pay into “a fund administered by the state” rather than paying the outside groups directly.
As we read it, the state route differs from Anthropic’s federal draft, which Politico reported on 16 October as a federal court in Washington, D.C. acting on a foreseeable, imminent risk of more than $1 billion in damage or more than 50 people killed or seriously injured.
Anthropic’s letter to Gov. Newsom, as quoted by Politico
“near-term risk of catastrophic harm”
Three Implications
FOR FRONTIER LABS Politico reports Anthropic says evaluators’ access “should be comparable to what a company’s own risk assessment staff have.”
FOR AUDITORS Politico reports state officials would rate auditors on criteria like “the rigor of their reasoning” and could require at least one highly rated evaluator.
FOR CALIFORNIA LAWMAKERS Politico reports the Legislature is expected to pursue a slew of regulations when it returns in about a month.
The framework describes Anthropic as “actively attempting to define the governance layer of the AI stack.”
As we read it, the Newsom letter is that attempt applied to a state: Politico reports Anthropic set out who could seek a stop, on what trigger, and how auditors are paid. The framework was written in June and does not discuss this letter or California.
What Is Not Established
We have not read Anthropic’s letter or OpenAI’s; Politico does not say they have been published, and Politico gives no exact date for Anthropic’s beyond “earlier this month.” Politico includes no response from Newsom’s office.
The reporting we read attaches no bill number or author to the letter’s proposals.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Politico reports that Anthropic has asked Newsom for a state route to suspend a risky model by court order, with auditors paid through a state fund. OpenAI, Politico says, backed shutdown controls and embedded auditors but not government power to block models.
A note on sourcing. The account of Anthropic’s and OpenAI’s letters to Gov. Newsom is Politico’s reporting; we have not read either letter, and every quote from them is as Politico reports it. The executive order details come from the Benton Institute’s summary. The proposals described are not law and are not attached to a bill in the reporting we read.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Who Pays the Auditors
