Senior Responsible AI Researcher

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

Research role on frontier AI safety, building evaluation frameworks, benchmarks and red-teaming methods to measure harms in generative AI models and agents. It sits in a safety company's research group and covers risk areas including hate speech, misinformation, elections, IP, and child safety, plus emerging risks like persuasion, manipulation, and sycophancy.

• Hate speech
• Misinformation
• Coordinated manipulation

What you'd do

• Build test sets and scoring methods that gauge what models can do, where they fail, and how well they hold to intended behavior
• Develop adversarial prompting and red-teaming strategies to stress-test LLMs, multimodal systems, and AI agents
• Study how models sway people, flatter users, show bias, or get jailbroken, across image, text, and agent setups
• Lead collaborative research projects with academic institutions and frontier AI labs
• Publish research findings and speak at technical AI safety conferences and industry events
• Translate research into safety products and client offerings with product, engineering, and policy teams
• Mentor junior researchers, data scientists, and engineers on technical work

What it asks for

• Doctorate in STS, ML, AI ethics, social data science, or a nearby discipline
• Strong publication record at top-tier ML, NLP, vision, or AI safety venues
• Significant hands-on time in AI safety, Responsible AI, Trust & Safety, or model evaluation
• Hands-on Python and modern ML frameworks such as PyTorch and Transformers
• Deep working knowledge of foundation models, agents, and safety mitigation techniques
• Experience designing datasets, benchmarks, red-teaming protocols, or simulation platforms
• Prior work with frontier AI labs or safety research collectives (preferred)

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.