About the job
About Anthropic
At Anthropic, we are dedicated to developing AI systems that are reliable, interpretable, and controllable. Our mission is to ensure that AI benefits individuals and society as a whole. We are a rapidly expanding team of passionate researchers, engineers, policy experts, and business leaders united in our efforts to create safe and beneficial AI systems.
About the Team
The Frontier Red Team (FRT) is a specialized technical research group within Anthropic's Policy division. Our mission is to enhance global safety in the age of advanced AI by thoroughly understanding the capabilities of these systems and developing effective defenses against potential threats.
In 2026, we are concentrating on research aimed at ensuring the safety of self-improving, highly autonomous AI systems, particularly those with cyberphysical capabilities. Explore our previous work on cyberdefense, robotics, and Project Vend. This is groundbreaking research with the potential for significant impact.
About the Role
As a Research Engineer on our team, you will tackle the critical challenge of defending against the potential adversarial use of powerful, autonomous, self-improving AI systems.
Your role will involve constructing and evaluating model organisms of autonomous systems and developing the defensive mechanisms necessary to counteract them. This work lies at the intersection of AI capabilities research, security, and policy—what we discover will directly influence how Anthropic and the wider world prepares for advanced AI.
This is applied research with substantial implications. Your contributions will inform decisions at the highest echelons of the company, aid in public demonstrations that shape policy discussions, and help develop technical defenses that could be crucial as AI systems evolve.
What You Will Do
- Design and construct autonomous AI systems capable of utilizing tools and operating in varied environments, creating model organisms that enhance our understanding and defenses against advanced adversarial AI.
- Develop evaluations and training environments to influence agent behavior in beneficial ways.
- Create defensive agents that can detect, disrupt, or outmaneuver adversarial AI systems in realistic scenarios.
- Integrate Claude with hardware platforms (e.g., robotics, physical systems) to assess cyberphysical risks and defenses.