Your Impact at LILA

Lila is developing AI systems for scientific discovery, connecting advanced models with scientific tools and automated laboratories. As these systems become more capable, we need to understand how they could be misused or compromised, which protections matter most, and how to establish that those protections work together.

We're hiring an experienced security practitioner to lead this work at the boundary between AI safety and security. You will investigate real systems, develop threat models, test important assumptions and recommend the protections needed for particular capabilities and deployments. Your work will address both dangerous misuse and the protection of sensitive models, data and research.

This is an opportunity to help shape a developing area of AI safety through security practice. You will combine technical investigation with careful analysis and clear writing, working closely with our established Security team, researchers and engineers to turn findings into implemented improvements.

You will bring security depth and develop your understanding of model capabilities, evaluations and safeguards through that work. Prior experience in AI safety is welcome, but not required.

What You'll Be Building

• Threat-model consequential systems and capabilities. Investigate how dangerous misuse or compromise could occur across models, deployed services and scientific workflows. Identify the relevant actors, access routes and failure paths, and the protections that could interrupt them.
• Test the assumptions behind our defenses. Examine architectures, code, configurations and operational evidence. Run focused adversarial tests, build prototypes and investigate weaknesses, including situations where several protections depend on the same underlying system.
• Shape the protections needed for safe deployment. Turn dangerous-capability assessments into concrete security requirements. Assess how compromise of models, sensitive assets or supporting systems changes the safety assessment, drawing on personnel, physical and other security specialists where relevant.
• Make evidence-based recommendations. Define what evidence would demonstrate that important controls and safeguards work together. Combine findings from model evaluations and security testing, explain the remaining uncertainty, and recommend improvements or conditions for use.
• Carry findings through to practical improvements. Work with Security, Safety and engineering owners to prioritize changes, specify acceptance criteria and verify the results. Make clear recommendations about which gaps require attention first and why.
• Anticipate what more capable systems will require. Identify where changes in capability, access or deployment call for stronger protections. Help teams plan security work with long lead times, and update assessments as systems change or new evidence emerges.
• Develop methods that others can use. Turn your investigations into clear threat models, assessment methods and technical guidance. Help Safety and Security colleagues apply the approach to new problems and build a shared understanding of the evidence behind decisions.

What You'll Need to Succeed

• Substantial hands-on security experience and deep expertise in at least one relevant discipline, such as security architecture, application or infrastructure security, offensive security, or incident response. You can reason across disciplines and identify when specialist judgment is needed.
• A record of resolving consequential security problems. You have investigated ambiguous risks, challenged designs and followed improvements through to validation, including work involving several teams.
• Strong technical judgment. You can examine unfamiliar systems, scrutinize code and configurations where relevant, and use technical investigation or testing to distinguish plausible claims from demonstrated protection.
• Curiosity about AI safety and scientific systems. You want to understand how model capabilities and safeguards are evaluated, can learn unfamiliar domains quickly, and reason carefully when evidence is incomplete.
• Clear writing and effective collaboration. You can explain a technical argument precisely, work through disagreements with specialists and turn a complex investigation into an actionable recommendation.

Bonus Points For

• Securing ML infrastructure, model assets, research environments or AI agents.
• Evaluating safeguards, adversarial robustness or defenses against sophisticated threats.
• Translating risk assessments into technical requirements, assurance evidence or deployment decisions.
• Working in scientific or other environments where digital access can have significant real-world consequences.

We value demonstrated security depth, sound judgment and the ability to learn. A research PhD or experience in every security discipline is not required.

Your early work will help establish how Lila combines safety and security evidence to make concrete decisions about its systems.