Google DeepMind is seeking to hire a nuclear expert to evaluate the risk that its frontier AI models could be used to aid weapons proliferation, according to a job listing posted this month in October.
The move signals that the AI lab is building an internal capability to directly assess how models such as large language models might uplift malicious actors in the nuclear domain, from fission weapons development to thermonuclear proliferation.
Automatica reported in October that DeepMind's Antigravity platform identified an antimicrobial peptide candidate, underscoring the laboratory's breadth of scientific projects. The nuclear expert hire, part of DeepMind's Chemical, Biological, Radiological, Nuclear, and high-yield Explosives (CBRNE) team, adds a distinct safety mandate to its research agenda.
The listing, for a Research Scientist role based in Mountain View, California, requires a PhD in plasma physics, nuclear engineering or a related field, and experience collaborating with US national labs or agencies like the IAEA, NNSA, or DARPA. Candidates should have a background in stockpile stewardship, nonproliferation, arms control, or threat reduction, and technical expertise in high energy density physics, inertial confinement fusion, implosion dynamics or reactor operations.
Preferred qualifications include experience in AI safety, red teaming, or applied machine learning research. The successful applicant will construct threat models prioritizing vulnerabilities from fission to thermonuclear weapons development, create evaluation datasets of complex physics prompts that elicit a model's maximum capability to assist malicious actors, and design mitigations that block harmful outputs without unduly degrading scientific utility, the listing states. The role requires building evaluations that are demonstrably unclassified and non-export controlled, using dual-use science proxies for weapons topics.
The position also involves supporting Google's nuclear safety policy and communicating risks to leadership and external stakeholders. Pay ranges from $174,000 to $252,000 per year in the US, plus a 15% bonus target, equity and benefits.
The listing does not mention a deadline for applications or when the position is expected to be filled.