AI Red Teamer, CBRNE (Remote)
Handshake · Remote · Remote (USA) ·
COMPENSATION
€•• – ••k
Listed on the employer's job board (Ashby)
- Fully remote position
- CBRNE domain expertise
- AI safety focus
- Adversarial testing
- Senior level role
Join Handshake AI to work at the intersection of AI safety and national security. You will use deep domain expertise to stress-test frontier AI models against chemical, biological, radiological, nuclear, and explosive threats, ensuring they do not provide actionable instructions for harmful activities.
What you'll do
- Design technically grounded adversarial prompts to test CBRNE threat handling
- Evaluate model outputs for technical accuracy and genuine danger levels
- Probe dual-use knowledge boundaries blending legitimate and weaponized uses
- Test multi-step attack chains simulating motivated actor behavior
- Score responses against structured harm taxonomies and severity rubrics
- Document findings with clear technical reasoning on model failures
- Distinguish between open literature info and operationally significant uplift
- Refine CBRNE-specific evaluation frameworks and threat models
- Collaborate with researchers to translate findings into model improvements
What you'll need
- Graduate-level education or equivalent in a CBRNE field
- Ability to assess technical accuracy and real-world consequences
- Understanding of dual-use research and open-source vs uplift distinctions
- Hands-on experience using multiple LLMs like ChatGPT or Claude
- Creative adversarial problem-solving skills
- Clear written communication for non-specialist audiences
- Strong ethical judgment separating adversarial thinking from values
- Self-directed and comfortable in feedback-heavy environments
Nice to have
- Active or prior security clearance (Secret, Top Secret, or SCI)
- Experience in threat assessment, WMD analysis, or intelligence
- Background in biosafety, nuclear nonproliferation, or EOD
- Familiarity with regulatory frameworks like CWC, BWC, or IAEA
- Experience in red teaming or structured adversarial evaluation
- Familiarity with Python, LLM APIs, or evaluation tooling
- Published research in a relevant CBRNE domain
- Prior work in trust and safety or AI evaluation
Stack
- ChatGPT
- Claude
- Gemini
- Python
- LLM APIs
How you'll work
- Fully remote role
- Global location flexibility
- Structured evaluation framework
- Strict ethical guidelines and security protocols