Project
AI does not need to become “superintelligent” to do harm. It just needs a state that holds its switch.
Client
Services
Feminist Analysis
Investigative Piece
AI Governance
Article
Year
2026
The Challenges
Introduction
By now you've seen it. An Anthropic researcher resigned saying the labs are "gambling with our lives." His alignment lead agreed that AI "could kill all humans." But that's not the end of the story. It's the middle.
our approach
What happened
On 9 September, researcher Jacob Coxon resigned from Anthropic and posted that the company and its rival OpenAI are “racing straight to self-improving superintelligence and gambling with our lives.” Within hours Anthropic’s alignment lead, Evan Hubinger, replied that Coxon was right. “We really do earnestly believe AI could kill all humans,” he wrote, and put the odds above ten percent within the decade. Colleagues at both labs followed. Samuel Marks, who leads Anthropic’s oversight research, added that “the more senior the employee, the more concerned they are.” An OpenAI researcher put the number at seventy percent absent regulation. Three days later Anthropic’s chief executive published an essay: “We must slow the pace at which we improve the capabilities of AI models.” OpenAI’s chief executive agreed within hours. By the weekend the Associated Press was reporting that Dario Amodei had warned a swarm of AI agents could take over the internet within six to twelve months. On the same day Coxon resigned, a committee of the UK Parliament formally agreed a report on human rights and the regulation of AI. Among its recommendations is that the development of systems “such as Artificial General Intelligence and Artificial Superintelligence, which risk causing widespread and very serious harm, including the capacity to evade effective human control, would be prohibited.” So for the first time, the people inside the labs and the people inside a parliament are saying roughly the same thing. Slow down. Test before release. Give someone the power to say no. That is the story that will be written about this month. This is the other one.
the result
The Ninety Days That Make This Urgent
Between June and September 2026, the power that every AI safety framework now proposes was exercised, rebuilt, and written into law, and the record shows for whom. On 12 June the US Commerce Department switched off two frontier models for the entire planet by letter, with no process, for eighteen days. On 15 June the Justice Department moved to protect the gas turbines poisoning Memphis-area Black neighbourhoods, calling the chatbot they power "a matter of paramount national security." On 11 August a watermark built for European consumers was applied worldwide without the rights framework that came with it. On 26 August a federal court entered a $17 billion Meta settlement that mandates age assurance on every user and hands state attorneys general a raw-data pipeline, in states that have already used Meta records to prosecute an abortion. On 9 September, lab insiders and a UK parliamentary committee reached the same conclusion on the same day, slow down and give someone the power to say no, without asking who that someone is or whose side they are on.
FIRST
the extinction frame and the rights frame are built on the same absence, one by dissolving the victim into "all humans," the other by scoping the Global South, data labourers, caste, and LGBTQ people out of a hundred-page report entirely
SECOND
the switch has already been thrown once, by one government, over the whole planet, to serve a security concern that belonged to no one who lost the tool, and the company that makes the model endorsed the switch and objected only to how it was used
THIRD
whether we are in the room or not, these frameworks are being built now, and the people they land on hardest are the ones they were written without, which is who needs to be where they are written
