Russell calls for AI 'red lines' over Amodei's 'pacing' plan

Altman, Musk, Hassabis, Nadella back Amodei letter

On September 15, 2026, The Guardian published an opinion column by Stuart Russell, a distinguished professor of computer science at the University of California, Berkeley, and president of the International Association for Safe and Ethical Artificial Intelligence. Russell wrote that the column was prompted by “a week of high drama in AI,” precipitated by the resignation of AI safety researcher Jacob Coxon from Anthropic and “several weeks of increasingly lurid and disturbing revelations about the OpenAI/Hugging Face incident.”

Russell’s inbox that day, he wrote, included a Business Insider message with the subject line “AI doomsday debate reaches boiling point.” The week’s central document, according to Russell, was a 3,800-word letter from Anthropic CEO Dario Amodei titled “We Must Pace the Frontier,” proposing a three-part framework for managing AI development. Russell wrote that OpenAI’s Sam Altman, Elon Musk of xAI, Demis Hassabis of Google DeepMind, and Microsoft’s Satya Nadella had all expressed support for the letter.

Amodei’s proposal has three parts, Russell wrote. The first calls for third-party AI system evaluators working inside each company with full access to the systems; Amodei committed Anthropic to the plan without waiting for the government to require it. The second asks frontier AI companies in “democratic countries” to “establish common safety standards as well as limits on the rate of unchecked AI progress,” with government regulation where needed. The third would include “authoritarian countries” in a broader compact. Russell wrote that Amodei’s letter “goes out of his way to reassure those in Washington who see America’s lead in AI as its most important geopolitical asset.”

The phrase “pacing the frontier” also appeared in a July open letter titled “Pacing the Frontier,” signed by 1,386 employees of frontier AI labs including Amodei, Russell wrote. That earlier letter noted “the complete absence of credible plans for controlling superintelligent AI systems” and asserted that “building things smarter than humans … is, objectively, an insane and suicidal thing to do.” Amodei’s letter, Russell wrote, was prompted by his concern that “AI has been advancing drastically faster, driven primarily by … recursive self-improvement.”

On a “casual reading” of Amodei’s letter, Russell wrote, “there are many reasons to believe that Amodei is calling for a general slowdown in the rate of progress.” Amodei invoked “limits on the rate of unchecked AI progress” and “some kind of ‘speed limit’ on the rate of recursive self-improvement (RSI),” according to Russell. Amodei wrote: “Progress will still seem fast, and we must make wise use of the time we gain.” Amodei also wrote: “To be clear, pacing does not mean halting model training or technical progress.”

Russell noted that Amodei talks about one to two years of extra time for research on interpretability, alignment, and better testing methods. Russell characterized the pacing metaphor as “completely misguided.” “We cannot set a slower rate of progress for capabilities and then hope that provides enough time to get the safety right,” Russell wrote. “The safety requirements are non-negotiable. We must set the safety requirements first, and further progress occurs only when they are met.”

Russell compared the alternative to Boeing announcing “a new plane every year” and hoping for time to complete flight tests. The correct sequencing, he wrote, requires that any new aircraft pass all tests and receive government airworthiness certification before it is introduced. “If that takes more than a year, so be it,” Russell wrote.

Russell dismissed as “nonsense” the criticism that Amodei’s letter is a bid to cement Anthropic’s lead through government intervention. The Wikipedia entry on pacing in F1 races, Russell wrote, describes the mechanism as one that “eliminates any time and distance advantage that a leading driver may have had over the remaining field of competitors.”

On a “more careful reading” of Amodei’s letter, Russell wrote, the Anthropic CEO agrees with the objection. Amodei wrote that rules should take the form: “If models have capability X, then they need to be accompanied by certifications of alignment properties Y and Z.” That formulation, Russell wrote, “is in fact the ‘red lines’ approach that AI safety researchers have been calling for.” The implication, he added, is that if developers cannot meet safety requirements, they must halt. “It would be, in F1 terminology, a red flag and not a pacing car,” Russell wrote.

Russell argued there is “no plausible alternative” to the red-lines approach because recursive self-improvement leading to superintelligent AI raises the risk of irreversible loss of human control. The acceptable risk level for loss of control is, he wrote, “perhaps one in 100m per year, not the one in 10 or one in five that the AI CEOs currently estimate.”

Russell wrote: “At some point, progress along this technology path will halt, not because further progress is impossible, but because further progress is untenable when the technology is intrinsically unsafe. Humanity has a right to protect itself.”

Russell acknowledged “huge resistance” to the conclusion that some AI development paths must end, citing trillions of dollars already invested in the current technology path and trillions more planned. “But the sunk cost fallacy is just that: if we double down on a mistake, it’s still a mistake,” he wrote.

The present level of attention to AI risk, the unanimity of leading technology executives, and a forthcoming Trump–Xi summit give “a real opportunity to choose a different path,” Russell wrote. “We must take it,” he concluded.

Russell is a Guardian US columnist, according to the column’s byline.