AI Red Team – GenAI Adversarial Evaluation

Written by the T&S Network from Alice's posting.

AI red team work on LLM and AI systems: designing and reviewing adversarial attacks such as indirect prompt injection, jailbreaks, agent manipulation and tool-use attacks, run in controlled and authorized environments. The posting places the role in a red teaming program that builds scalable attack generation and test systems.

• AI misuse and jailbreaks

What you'd do

• Create novel adversarial attacks against LLM and AI systems.
• Help build scalable systems that generate and test attacks at large volume.
• Review attacks for quality, effectiveness, novelty and reproducibility.
• Spot emerging attack patterns and weak points in models.
• Develop adversarial datasets, benchmarks and regression tests.
• Share findings, techniques and structured attack data with Data Science, Security and Engineering teams.
• Help strengthen model robustness and defensive measures.

What it asks for

• Three or more years of professional work in cybersecurity, security research, red teaming or a nearby field.
• Solid grasp of LLMs, prompt injection and AI security.
• Experience with red teaming, security research or adversarial testing.
• Proficient in Python and scripting automation.
• Inventive problem-solving skills and a knack for devising new attack methods.
• Can interpret test results and explain findings clearly.

Terms in this role

Red teaming.

Testing a system, often an AI model, by deliberately trying to make it fail or misbehave, so problems are found and fixed first.

Jailbreak.

A prompt written to get an AI model to ignore its safety rules.

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on Alicenorth_east before you apply. If we got something wrong, report this listing.