U.S. artificial intelligence pioneer Anthropic is seeking a specialist in chemical weapons and high-yield explosives to protect its technology from potentially disastrous abuse.
The LinkedIn job advert calls for at least five years’ experience in such defence areas, along with knowledge of radiological dispersal devices, or dirty bombs. At its core, the firm worries its AI could guide malicious users in building chemical or radioactive arms, so it wants robust barriers in place.
Anthropic assured the BBC that this position echoes roles it has established in other high-risk domains. The company is not alone. OpenAI, behind ChatGPT, advertises for a researcher on biological and chemical threats, promising salaries up to $455,000—nearly twice Anthropic’s offer, as Reuters has reported.
Such moves alarm some observers, who argue that inputting weapons details into AI systems risks leaks, despite prohibitions. Tech researcher Dr Stephanie Hare, who co-presents the BBC’s AI Decoded, raised pointed concerns.
“Is it ever safe to use AI systems to handle sensitive chemicals and explosives information, including dirty bombs and other radiological weapons?” she asked. “There is no international treaty or other regulation for this type of work and the use of AI with these types of weapons. All of this is happening out of sight.”
These hires gain urgency as the U.S. presses AI firms during its war against Iran and actions in Venezuela. Anthropic is locked in a lawsuit with the Pentagon after a supply chain risk label, triggered by its refusal to support autonomous killers or mass U.S. surveillance. Co-founder Dario Amodei noted in February the tech’s unreadiness for such roles, prompting White House claims that military autonomy prevails—much like Huawei’s security-driven blacklist.
OpenAI aligned rhetorically but landed its own paused government deal. Claude, Anthropic’s AI, still powers Palantir tools in the U.S.-Israel Iran fight. With no binding global pacts akin to the 1993 chemical weapons treaty, and stalled UN talks, self-policing dominates amid existential warnings from pioneers like Geoffrey Hinton.