Cyber Evaluations Engineer at Anthropic

Confirmed open

Design cyber capability and safety evaluations for Anthropic AI models, test safeguards before releases, and build Python-based tools and misuse detection probes.

Anthropic

Location: Remote-Friendly, United States; San Francisco, CA | Washington, DC / San Francisco, California, United States
Work arrangement: remote
Technologies: Python

Role overview

Anthropic is hiring a Cyber Evaluations Engineer to assess cyber capabilities and the robustness of safeguards in its AI models. The role combines model evaluations, analysis of jailbreak and prompt-bypass data, and design of probes for detecting misuse.

Responsibilities

Design and run cyber capability, uplift and safety evaluations for new models.
Test safeguard robustness before major model releases and explain the results to stakeholders.
Prototype and tune production probes for cyber misuse, then measure their precision and coverage with the cyber policy team.
Build internal tools to run and score evaluations and turn findings into safeguard improvements.

Qualifications

Experience building or running evaluations, benchmarks or test suites for software or machine-learning systems.
Hands-on cybersecurity work such as CTFs, vulnerability research, exploit development or security research.
Python proficiency and the ability to communicate technical results across teams.
The posting also values offensive-security research, adversarial-data analysis and AI evaluation frameworks.

Benefits

The posting lists vacation and parental leave, flexible working hours and optional equity donation matching.
Anthropic states that staff generally spend at least 25% of their time in an office; some roles require more. Check the posting for this role’s location and travel conditions.

About Anthropic

Anthropic develops AI systems with a focus on reliability, interpretability, steerability and safety.

How to apply

Review the original posting and apply through the employer’s careers page.

View source and apply

Source checked: 2026-09-30

Show what you can do

Build a free profile around your CV and real work. Share its link when you are ready. Applications for this role still go directly to the employer.

Build a free profile

Browse job categories

Python