SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 300,000 - 405,000 / annual
Anthropic is hiring Cyber Evaluations Engineers to design and execute evaluations that measure cyber-relevant capabilities and safeguard robustness in AI models. In this role, you will build and run capability, uplift, and safety evaluations to assess cyber-relevant risks in new models ahead of major releases. You'll analyze evaluation results and communicate findings to cross-functional teams and stakeholders, then work with policy and engineering partners to translate those findings into safeguard improvements.
Key responsibilities include designing, prototyping, and tuning detection probes for cyber misuse; executing per-release safeguard-robustness testing; and collaborating with the cyber policy team to develop a layered, robust abuse-detection architecture. You'll also build and maintain internal tooling used to run and score evaluations, and analyze data on jailbreaks and prompt bypasses to understand where safeguards hold up and where they don't.
Minimum qualifications include hands-on experience building or running evaluations, benchmarks, or test suites for software or ML systems with delivery on fixed timelines; practical cybersecurity experience such as CTF participation, vulnerability research, exploit development, or security research; proficiency in Python; and strong communication skills for presenting results to multiple stakeholders.
Preferred qualifications include deep offensive-security or security-research experience with AI security benchmarks; experience analyzing adversarial or abuse data; familiarity with AI/ML evaluation frameworks; experience authoring detection content (Sigma, YARA, Suricata, SIEM rules) or building ML-based abuse detection; and active secret security clearance or eligibility to obtain one. Experience working with government partners on testing engagements is also valued.
