Head of AI Safety

Company overview

DeepRec.ai is partnered with a well-funded, fast-growing AI security company to find their first Head of AI Cyber Safety.

They help frontier AI labs and large enterprises deploy AI safely. Their work covers adversarial testing of large language models, automated red-teaming of AI agents, and real-time protections that detect and block misuse in production. Customers include several leading AI labs and global enterprises.

The company has strong commercial traction, a recent major funding round and an established engineering and research team. This hire will build a brand-new function inside a business that already has the scale and customers to make it count.

The role

As Head of AI Cyber Safety, you'll build and lead my client's cyber safety capability from the ground up. You'll be their technical authority on AI-enabled cyber risk, across red-teaming, evaluations, benchmarks and defensive tooling.

The role blends hands-on offensive security with AI safety and ML engineering. You'll turn deep attacker know-how into repeatable evaluations, datasets and automated defences that customers rely on, and that help set the standard for the wider industry.

You'll report to the CEO and work closely with the research, ML engineering and customer teams. Expect to stay close to the technical work while you hire and lead a team of specialists.

What you'll do

• Run adversarial cyber evaluations. Design and lead testing of frontier models and AI agents for offensive capability: vulnerability discovery, exploit development, malware authoring, privilege escalation, lateral movement, social engineering and autonomous intrusion, across chat, agentic and multimodal settings.
• Turn expertise into infrastructure. Partner with the ML engineering team to build benchmarks, test harnesses, cyber ranges, classifiers and guardrails that run at scale, both in the company's own products and in customer deployments.
• Own the risk taxonomy. Create and maintain the company's framework for severe and catastrophic cyber harm, and keep it current as model capabilities move.
• Advise the people deploying AI. Write clear technical risk assessments and mitigation recommendations for AI developers, enterprise security teams and the leadership team.
• Build the team. Hire, mentor and lead a team of offensive security specialists, and set the standards, processes and tooling they work with.
• Be the company's voice in the field. Represent the business with AI labs, security researchers, public-sector partners and the AI safety and security communities, through collaborations, talks and published research.

Who you are

• You have deep, hands-on offensive security experience: vulnerability research, exploit development, penetration testing or red teaming, malware analysis or reverse engineering, gained in industry, government, academia or on your own.
• You understand how real attackers operate, ideally including threats to critical infrastructure or other high-consequence environments.
• You have tested AI systems adversarially: red-teaming LLMs, LLM security research, or building evaluation sets for capable models.
• You can move comfortably between security research, AI safety and ML engineering, and explain each to the others.
• You do your best work with ambiguity, setting direction while building something that doesn't exist yet.
• You like building: teams, tooling and processes from a blank page.

Don't meet every point? Get in touch anyway. We care far more about what you can do than a checklist.

Nice to have

• Experience building ML models, security classifiers or automated detection systems.
• Hands-on jailbreaking, prompt injection research or evaluation of tool-using AI agents.
• Work with frontier AI developers, national security bodies or top-tier security research teams.
• A background in threat intelligence, autonomous cyber operations, agent security or AI governance.
• Strong engineering skills in Python, Go, Rust or another systems language.
• Public research, CVEs, CTF results, conference talks or open-source tools you can share.

Why this role

• A function to build, not inherit. You'll define how the company approaches cyber safety, with senior backing and the budget to hire.
• Work on the frontier. You'll test some of the most capable AI systems in the world and see how they behave under real adversarial pressure.
• Impact you can see. Your evaluations will shape how AI is deployed, and your defences will run in production for major customers.
• Scale with stability. You get the reach of an established, well-funded company with the pace and ownership of a growth-stage one.
• A mission that matters. Help make sure increasingly capable AI makes defenders stronger, not attackers.

What's on offer

• Base salary: $350.000 - $450.000 plus a performance bonus
• Equity: a meaningful equity package
• Flexibility: hybrid or remote
• Growth: a learning and conference budget, and time to publish research
• Visas: sponsorship available for exceptional candidates