Job description
google.com • Mountain View, CA, USA; New York, NY, USA; Zürich, Switzerland
San Francisco Bay Area, New York, NY, Zurich, Switzerland
$174,000 - $253,000
- In this role, you'll conduct security assessments on agentic AI services, identifying vulnerabilities like prompt injection and tool-use escalation.
- Engineer and execute multi-turn attack sequences exploiting model behaviors, logic errors, and data poisoning vectors.
- Develop automated regression testing frameworks from manual vulnerability discoveries.
- Partner with developers during design phases to provide immediate feedback on security findings.
- Maintain and expand a library of agent-specific attack patterns and exploit primitives.
Google DeepMind is a frontier AI research and product company, with teams working on alignment, policy, and security. We post specific opportunities at Google DeepMind that we think may be high impact. We do not necessarily recommend working at other positions at Google DeepMind. You can read concerns about doing harm by working at a frontier AI company in our career review on the topic.
Applications are handled by the employer.