AI Safety Research Scientist: Agent Misbehavior & Alignment

Aisafety; Paris; Hybride; EUR 70 000 - 110 000; Plein temps

Avantages offerts par ce poste

Paid time off

Relocation package

Medical insurance (France)

Hardware & tools provided

Subscriptions for AI agents and IDEs

Team off-sites

Résumé du poste

White Circle is seeking a research scientist to study how LLM agents fail in real-world settings, eliciting deception and unsafe behaviour in concrete experiments. You will build understanding of how agents misbehave with realistic user interactions and develop evals for our products.

You will publish findings and contribute to internal guardrails, working in a small, focused team with opportunities to ship notable safety improvements.

Qualifications

• A track record of empirical research in agent behaviour, model evaluation, alignment, or adjacent area.
• Strong ML engineering: independently build a research MVP involving fine-tuning, agent inference, and evals.
• Experimental design under real conditions: isolating agent failure modes, calibrating judges and baselines.
• You can take a vague behavioural question and define the experiment that answers it, then run it fast.
• An AI power-user—fluent with frontier models and coding agents in daily work.

Responsabilités

• Own research projects end to end—from unclear concerns to falsifiable experiments and defend results.
• Develop automated audit agents that discover and characterize suspect model behaviour at scale.
• Study misalignment and bias when real users interact with agents and turn findings into evals.
• Pressure-test frontier agents in realistic, high-stakes scenarios to find where they break.
• Run white-box and black-box investigations to understand how AI models fail.
• Publish what you learn as public blog posts and conference papers, feeding back into guardrails.

Connaissances

Empirical research

ML engineering

Experimental design

Frontier models

AI power-user

Formation

MSc or PhD in ML/CS/cognitive science/comp neuroscience/physics

Outils

NLAs

SAEs

Description du poste

White Circle is seeking a research scientist to study how LLM agents fail in real-world settings, eliciting deception and unsafe behaviour in concrete experiments. You will build understanding of how agents misbehave with realistic user interactions and develop evals for our products.

You will publish findings and contribute to internal guardrails, working in a small, focused team with opportunities to ship notable safety improvements.