Anthropic’s job ads read like a threat assessment

A look at Anthropic safety hiring shows exactly what it fears: analysts brought in to stop its models teaching anyone how to build nuclear, chemical, and biological weapons.

Most job ads sell a mission. Anthropic’s read like a threat assessment.

The company has posted a run of openings for enforcement analysts whose job is to keep its AI from helping people build weapons, run scams, or commit cybercrime, Axios first reported. One listing seeks an “Enforcement Analyst focused on Radiological & Nuclear Harms.” Others cover chemicals and explosives, financial fraud, and more.

The pay lands in the mid- to upper-$200,000s. The work is not coding. Anthropic wants real-world expertise in fields like biology and explosives. It also wants people who can think like an attacker trying to slip past its defences.

Naming the harm on purpose

The blunt job titles are deliberate. “Ensuring our models don’t provide potentially harmful information is central to responsible development,” a spokesperson said. The company said it regularly hires experts in sensitive fields to stress-test its models before a release.

Spelling out the exact harm, it added, is how you recruit the right people. Anthropic says hundreds of staff now work on safety, probing for weak spots and patching them.

This is the company that critics call the industry’s biggest doomsayer. The pattern in Anthropic safety hiring is its answer to that label. It is spending real money on the risks it keeps describing.

The catastrophe Amodei keeps describing

Chief executive Dario Amodei has spent months sketching the downside. In a January essay he called biological attacks the scenario that worries him most.

“I do not think biological attacks will necessarily be carried out the instant it becomes widely possible,” he wrote. “But added up across millions of people and a few years of time, I think there is a serious risk of a major attack, with casualties potentially in the millions or more.”

He has also warned about AI helping cybercriminals and empowering authoritarian states. Earlier this year Anthropic broke with the US Defense Department over the use of its technology for mass surveillance and autonomous weapons.

The labs are writing their own rules

OpenAI is doing the same. It is hiring a researcher on biological and chemical risks, at a base salary of up to $445,000. As models grow more capable, every serious lab is racing to staff a red team.

That race is happening in a vacuum. The US still has no comprehensive AI safety law. Congress has tried for years and passed nothing. Some want a referee: Google’s Demis Hassabis has floated a Wall Street-style watchdog for frontier models. Fewer than one in a hundred AI PhDs go into government, so the expertise sits inside the companies.

The result is a strange kind of self-regulation. The firms building the most dangerous capability are also the ones deciding how to fence it in. Amodei has named that tension himself, calling AI companies the next tier of risk after hostile states. His careers page is the argument and the warning in one place. The people best placed to stop the catastrophe work for the company that could help cause it.