Role Summary
Measure cyber-related model capabilities and safeguard robustness. Evaluate jailbreaks and misuse signals to identify risks before launches and improve protective controls.
This published opportunity is available through Toloka AI. Review the listing carefully for the specific work arrangement and requirements.
Responsibilities
Design capability, uplift, and safety evaluations. Run safeguard-robustness tests before releases. Analyze results and explain findings. Prototype cyber-misuse detection probes. Maintain evaluation and scoring tools.
Qualifications
Software/ML evaluation experience; hands-on cybersecurity experience; Python; clear cross-functional communication; relevant bachelor's degree or equivalent education, training, or experience.
Preferred Qualifications
Offensive-security research, AI benchmarks, adversarial-data analysis, government testing, detection content, or security-clearance eligibility.
Benefits
Vacation and parental leave; optional equity donation matching.
What happens next
Submit your interest online to receive a reference number. The opportunity owner or review team may contact selected applicants about next steps.
For accommodation requests, contact [email protected].