Anthropic hiring weapons experts to guard AI from misuse

Anthropic is on a hiring spree for enforcement analysts and subject-matter experts tasked with preventing its AI systems from being used to help build weapons of mass destruction, commit cyberattacks, or carry out financial fraud.

The company has posted a series of job listings in recent days seeking specialists in radiological and nuclear harms, chemicals and explosives, cybercrime, and other threat categories, as first reported by Axios and detailed by Yahoo Finance. The roles, which offer salaries in the mid- to upper-$200,000 range, do not require coding skills — Anthropic instead wants real-world expertise in fields like nuclear physics, biology, and explosives, along with the ability to think like an attacker probing for weak spots in the company's defenses.thenextweb+1

A Growing Safety Apparatus

One listing for a "Safeguards Enforcement Analyst, Radiological & Nuclear Harms" calls for candidates with backgrounds in nuclear engineering or health physics and experience in WMD nonproliferation or threat intelligence. Another, focused on cyber harm, asks analysts to detect and mitigate attempts to weaponize Anthropic's models for malicious cyber operations. The company is also hiring biological safety research scientists with PhDs in virology, microbiology, or related fields to "define what responsible AI safety looks like in the biological domain".job-boards.greenhouse+2

An Anthropic spokesperson told The Next Web that "ensuring our models don't provide potentially harmful information is central to responsible development," adding that hundreds of staff now work on safety. The BBC reported earlier this year that Anthropic was seeking a chemical weapons and high-yield explosives expert to mitigate what it called "catastrophic misuse" of its technology.bbc+1

Amodei's Repeated Warnings

The hiring push tracks closely with warnings from CEO Dario Amodei, who has spent much of the past year raising alarms about the risks posed by advanced AI. In a January essay, Amodei wrote that biological attacks worried him most, noting that "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 of a six- to 12-month window for companies and governments to patch software vulnerabilities that Anthropic's latest models have identified.axios+2

Self-Regulation in a Legislative Vacuum

Anthropic is not alone in this effort. OpenAI is also hiring researchers focused on biological and chemical risks, at salaries of up to $445,000. The parallel push underscores an unusual dynamic in the AI industry: the companies building the most powerful models are also the ones staffing up to police them, in the absence of comprehensive federal AI safety legislation. Amodei has acknowledged the tension, calling AI companies themselves a tier of risk just below hostile states.thenextweb