How to apply for Head of Cyber Safety

About Gray Swan

Gray Swan is an AI security company that builds tools to automatically assess the risks of AI models. The company focuses on evaluating frontier LLMs for dangerous capabilities, including offensive cyber use. Working here means being at the center of AI safety and cyber risk at a time when both fields are converging.

About the role

The Head of Cyber Safety will build and lead Gray Swan's cyber safety capability as the technical authority on AI-enabled cyber risk. The role involves designing adversarial evaluations of frontier LLMs for offensive cyber capabilities such as vulnerability discovery and exploits. You will partner with ML engineers to turn cybersecurity expertise into scalable benchmarks and detection systems, while developing cyber harm taxonomies and leading a team of cybersecurity experts.

A typical day

A typical day might involve designing adversarial evaluations for a new LLM, meeting with ML engineers to turn those evaluations into benchmarks, and mentoring team members on evaluation processes. You may also review emerging research on AI cyber risks and update harm taxonomies or detection systems accordingly. Since the role is remote, much of the work happens through written communication and virtual collaboration.

Who Gray Swan is looking for

• Deep hands-on experience in cybersecurity, especially offensive security, vulnerability discovery, or exploit development.
• Familiarity with frontier LLMs and how they can be evaluated for dangerous cyber capabilities.
• Experience building or leading technical teams, with the ability to mentor cybersecurity experts and establish evaluation processes.
• Ability to work cross-functionally with ML engineers to translate security expertise into scalable benchmarks and detection systems.

Tips for this application

• Highlight specific projects where you assessed or tested AI models for cyber risks, even if informal or research-based.
• Show experience designing adversarial evaluations or red-teaming exercises, especially for LLMs or other AI systems.
• Demonstrate leadership by describing teams you have built or mentored, and processes or infrastructure you have established.
• Explain how you have translated cybersecurity knowledge into scalable tools, benchmarks, or detection systems.
• Reference Gray Swan's focus on automatic risk assessment of AI models and connect your experience to that specific mission.

What to cover in your cover letter

Emphasize your technical authority in AI-enabled cyber risk. Describe concrete examples of designing adversarial evaluations for LLMs. Explain how you have partnered with ML engineers to build scalable benchmarks or detection systems. Outline your approach to building and leading a cyber safety team, including developing taxonomies and evaluation frameworks.

Research before applying

• Read Gray Swan's website and any public materials about their tools for automatically assessing AI model risks.
• Look into recent research or publications on adversarial evaluations of LLMs for cyber capabilities.
• Understand common cyber harm taxonomies and evaluation frameworks used in AI safety.
• Check if Gray Swan has published blog posts, papers, or talks about their approach to AI security.

Likely interview topics

Based on the job description, expect questions about:

• How would you design an adversarial evaluation to test a frontier LLM for offensive cyber capabilities like vulnerability discovery?
• What cyber harm taxonomies do you think are needed for AI models, and how would you maintain them as capabilities advance?
• Describe a time you translated cybersecurity expertise into a scalable benchmark or detection system. What was the outcome?
• How would you build and mentor a team of cybersecurity experts focused on AI safety evaluations?
• What are the key challenges in evaluating AI models for cyber risk, and how would you address them at Gray Swan?

Common mistakes to avoid

• Focusing only on traditional cybersecurity without connecting it to AI or LLM evaluation.
• Claiming leadership experience without concrete examples of building teams, processes, or infrastructure.
• Ignoring the need to collaborate with ML engineers or failing to show how you translate security expertise into scalable systems.