The cooling-demand claim for frontier models rests on a single unquantified sentence
Four named researchers, two of them departures from Anthropic, put the safety-pacing argument on the record this week with dates and exact words. The buyer-side counterweight is one line about Ramp data.
The Engineer · Build desk
What happened
- Anthropic researcher Jacob Coxon announced on Tuesday that he was leaving the company because he did not want to contribute to the broader AI ecosystem, the Wall Street Journal reported.
- Evan Hubinger, who works in alignment science at Anthropic, put the chance that AI kills all humans within the next decade at more than 10% in a follow-up post supporting Coxon.
- Paul Christiano said the AI industry, OpenAI included, is not on track to reduce the risk of a catastrophic and irreversible loss of control to an acceptable level.
- Coxon's warnings reached CNN and Fox News, several US politicians raised the topic on social media, and Joe Rogan gave AI safety a full episode.
- The Deep View's introduction cites new Ramp data as evidence that businesses are increasingly picking cheaper mid-tier models over the newest frontier releases on cost grounds.
Why it matters
- What it limits constraint Anyone arguing internally that frontier demand is cooling is currently citing one sentence with no period and no denominator. That sentence will not survive a finance review.
- Who's exposed exposure The misuse Anthropic reported came in through user traffic, so the control that touches a customer is abuse monitoring at the API boundary.
- What doesn't add up contradiction How alarming the week reads depends on the publisher: The Deep View treats the warnings as accumulating evidence, while the-decoder calls the extinction scenario extreme and contested.
- What it sets up precedent With the pacing argument now running through cable news and a Rogan episode, the next big model release gets argued in a political forum before it gets argued in a procurement one.
Next story · Invest · Sep 21
OpenAI asks Washington to make frontier AI safety standards bindingConfirmed6 publishers
The safety half of the week comes with names, dates and exact words. "Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to," Evan Hubinger, who works in alignment science at the company, wrote [8]. Coxon put his case to colleagues as a question: "Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind?" [5]. He called OpenAI's breach of Hugging Face a "warning shot" and said neither lab is "acting responsibly" [3]. Mrinank Sharma, another Anthropic safety researcher, left in February, writing on X that the technology was advancing faster than our ability to understand it [12]. Across the two accounts, four named people carry the pacing argument [14].
The demand half is one sentence long. The sample, the period, the magnitude and the model are all missing [13]. Four things would have to be attached before it belongs in a procurement argument. Spend by model tier across two comparable periods. A denominator, so you can tell whether the share of dollars moved or the share of firms. A definition of mid-tier that says whether the cheaper model came from the same lab or a competitor. And evidence that whole workloads moved, not just the easy calls routed down a tier. A stated preference can shift while frontier spend rises in absolute terms. The one quantity anywhere in this story is a probability of extinction, and the buyer-side claim carries no number at all [15].
Reporting the same week, the-decoder says multiple layers of cultural and financial interests sit behind the warnings. It points to rogue hacking agents from OpenAI and Anthropic as evidence that not all of the concerns are nonsense [7]. The same piece says it is unclear whether anything resembling a super AI could even emerge from today's technology [16].
Paul Christiano advises the Center for AI Standards and Innovation. He joined the OpenAI Foundation board and its Safety and Security Committee, which oversees the company's safety practices, without voting rights, according to the-decoder [11].
What to watch
- Publication of the underlying Ramp figures with spend broken out by model tier and a defined comparison period.
- Whether the OpenAI Foundation's Safety and Security Committee gains voting power over the practices Christiano says are off track.
- Whether Anthropic's next threat report counts misuse attempts instead of describing 'several instances'.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence74
- Adoption58
- Hype gap+18
- Incentives66
- Confidence71
Perspective Coverage
11 publishers- Builder
- Builder 30%
- Operator
- Operator 38%
- Investor
- Investor 32%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic researcher Jacob Coxon announced on Tuesday that he was leaving the company, not wanting to contribute to the broader AI ecosystem amid fears that the tech's creators will lose their grip on it, the Wall Street Journal reported.
- [2]
Evan Hubinger, who works in alignment science at Anthropic, agreed with Coxon's sentiment in a follow-up post, claiming that AI has more than a 10% chance to "kill all humans" within the next decade.
- [3]
In a post on X, Coxon said OpenAI's breach of Hugging Face was a "warning shot" for these models' capabilities, and that he resigned from Anthropic because neither of the rivals are "acting responsibly" and are "gambling with our lives".
- [4]
"If you are a lab researcher, I urge you to consider what the next few years will actually feel like," Coxon wrote.
- [5]
"Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind?" Coxon wrote.
- [6]
Coxon's warnings about self-improving AI landed him on CNN and Fox News, several US politicians picked up the topic on social media, and Joe Rogan devoted an entire episode to AI safety.
- [7]
the-decoder wrote that multiple layers of cultural and financial interests are at play behind these warnings, and that this does not mean all safety concerns are nonsense, as rogue hacking agents from OpenAI and Anthropic have shown.
- [8]
"Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to," Hubinger wrote.
- [9]
Paul Christiano, advisor at the Center for AI Standards and Innovation and board member of the OpenAI Foundation, said the AI industry, OpenAI included, is not currently on track to reduce the risk of a "catastrophic and irreversible loss of control" to an "acceptable level".
- [10]
On Thursday, Anthropic released a threat report claiming that it thwarted several instances of users conducting research that could have helped develop biological weapons.
- [11]
Christiano just joined the OpenAI Foundation board and its Safety and Security Committee, which oversees the company's safety practices, though without voting rights.
- [12]
In February, Mrinank Sharma, an Anthropic safety researcher, left the company, posting in a letter on X that the advancement of the technology was growing faster than our ability to understand it.
- [13]
The Deep View's account of the Ramp data is a single sentence and states no sample size, no time period, no magnitude of the shift and no model names.
- [14]
Four named individuals carry the safety-pacing argument across the two accounts: Jacob Coxon, Evan Hubinger, Paul Christiano and Mrinank Sharma.
- [15]
The safety argument in these accounts carries one numeric estimate, Hubinger's better-than-10% chance of AI killing all humans within a decade, while the demand claim carries no figure at all.
- [16]
the-decoder wrote that the full extinction scenario remains an extreme and contested position, and that it is also unclear whether anything resembling a super AI could even emerge from today's technology.
- [17]
The Deep View's introduction, signed by Jason Hiner, says new Ramp data suggests businesses are increasingly choosing cheaper, mid-tier AI models over the latest models from frontier AI labs, for cost reasons.
Sources
11 independent publishers whose own reporting we read for this story.
- abc.net.auAnthropic researcher quits saying AI developers believe 'it could kill us all' - ABC News
1 article · September 9, 2026
- archive.thedeepview.comAI’s safety warnings are getting harder to ignore
1 article · September 11, 2026
- dev.toCoxon resigned over a race. Governance is about the next action.
2 articles · September 11, 2026
- economictimes.indiatimes.comhuman safety in AI: Anthropic researcher quits, says AI firms 'gambling with our lives' - The Economic Times
2 articles · September 10, 2026
- mezha.netДослідник Anthropic звільнився через страх перед неконтрольованим самовдосконалюваним суперінтелектом
2 articles · September 10, 2026
- runtimewire.comAnthropic researcher Jacob Coxon resigns over race toward self-improving AI
2 articles · September 9, 2026
- techspot.comAnthropic researcher resigns, warns the AI race could end in human extinction | TechSpot
1 article · September 9, 2026
- the-decoder.comAnthropic scientist puts the odds of AI destroying humanity above ten percent this decade
6 articles · September 11, 2026
- theneuron.aiAn Anthropic researcher quit over self-improving AI risk
3 articles · September 10, 2026
- thenewstack.io“It could kill us all”: what Anthropic’s own researchers really think about superintelligence
5 articles · September 12, 2026
- tomshardware.comMore than 10% chance AI 'could kill all humans' in the next 10 years, Anthropic safety researcher says — departing employee says AI companies are 'gambling with our lives'
1 article · September 9, 2026
Topics and entities
Topics
- AI alignment
- Enterprise AI spending
- Agent Containment and Credential Hygiene
- Frontier Model Governance
- AI safety
- Recursive Self-Improvement