As AI systems evolve from tools into autonomous agents, they expose entirely new risk surfaces—emergent behaviors, agentic autonomy, and previously unimaginable vulnerabilities. Take the recent exploit in Cursor, an AI-native IDE, where a third-party attacker was able to manipulate the LLM via the rules file—resulting in fully compromised code. Or the LLM-based attack that exfiltrated private messages from Slack channels. These are just the beginning. The landscape is full of vulnerabilities waiting to be discovered, and this role puts you at the forefront of finding them—shaping both Lakera’s product direction and broader industry thinking.

At Lakera, we’re building security foundation models to craft highly effective adversarial attacks against LLMs. But it's not just about generating attacks in theory—they must be delivered in real-world systems to create real-world impact. This role bridges that gap between cutting-edge AI capabilities and practical exploitation.

We're looking for a Security Researcher with deep offensive expertise—someone who understands how systems break and is excited to apply that knowledge to the rapidly emerging domain of AI-native threats. Your work will directly shape how some of the largest and most advanced AI deployments in the world are tested, hardened, and trusted at scale.

About Lakera

Lakera is on a mission to ensure AI does what we want it to do. We are heading towards a future where AI agents run our businesses and personal lives. Here at Lakera, we're not just dreaming about the future; we're building the security foundation for it. We empower security teams and builders so that their businesses can adopt AI technologies and unleash the next phase of intelligent computing.

We work with Fortune 500 companies, startups, and foundation model providers to protect them and their users from adversarial misalignment. We are also the company behind Gandalf, the world’s most popular AI security game.

Lakera has offices in San Francisco and Zurich.

We move fast and work with intensity. We act as one team but expect everyone to take substantial ownership and accountability. We prioritize transparency at every level and are committed to always raising the bar in everything we do. We promote diversity of thought as we believe that creates the best outcomes.

Example Projects

Leverage Lakera’s internal security foundation models to discover and weaponize vulnerabilities in agentic systems.
Collaborate with AI researchers to uncover novel classes of LLM vulnerabilities—advanced jailbreaks, multi-step prompt injections, fine-tuning exploits—and use them to compromise real-world applications.
Design and lead red-teaming operations against internal and customer AI stacks, simulating adversaries to probe how LLMs fail in production—from misalignment to full system compromise.
Develop AI-native security benchmarks that reflect real exploitability, not just academic risk, and contribute to the first generation of practical evaluation standards for LLM defenses.
Work closely with ML engineers to scale offensive techniques into automated testing pipelines and embed detection capabilities into Lakera’s core products.
Shape the narrative of AI security—by publishing research, contributing to tooling, and helping define how traditional security paradigms must evolve in the face of intelligent systems.

About You

You are creative, bold, and ready to challenge assumptions. You are excited to tackle real-world AI security problems and see your research come to life. You enjoy working in a tight knit team that rapidly moves between ideation and implementation. You want to work in a fast moving team where you have ownership, impact, and direct influence on the secure deployment of agentic systems at scale.

We’re looking for someone with hands-on experience in offensive security, particularly in red teaming, penetration testing, or vulnerability research. You’ve found real issues in real systems, and you know how to think like an adversary.

In addition, any of the following would be valuable:

Strong engineering skills and the ability to build your own tools and infrastructure.
Familiarity with how modern machine learning systems work—or the ability to learn fast.
Experience with or interest in the security implications of LLMs and autonomous agents.
A track record of impactful security research, tooling, or public contributions.

If you’re ready to work at the frontier of AI security and truly make an impact, let’s talk.

👉 Let's stay connected! Follow us on LinkedIn, Twitter & Instagram to learn more about what is happening at Lakera.

ℹ️ Join us on Momentum, the slack community for AI Safety and Security everything.

❗To remove your information from our recruitment database, please email privacy@lakera.ai.

Research jobs in Zürich

AI Security Researcher (Zürich)