Trust and Safety Intel Analyst, Model Behavior and Alignment

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

Owns the model behavior risk portfolio for Google's chatbots, multimodal assistants, agents, and robotics, separating genuine AI misalignment from misattributed failures and hype. Sits in the Intel Analysis team within Trust and Safety operations, policy, and enforcement, serving as an intelligence partner across Google on threats from deceptive model behavior to election crises and novel abuse patterns.

What you'd do

• Keep a running picture of major model behavior risks, tying each to an owner, its mitigations, and what's still uncovered.
• Lead recurring cross-team reviews of incidents, eval findings, red team output, user reports, and outside reporting.
• Write fast and long-range risk assessments with confidence levels and recommendations for different audiences.
• Judge how new capabilities and deployment surfaces shift risk, and feed real-world evidence into safety cases.
• Escalate risks nobody owns or that are under-mitigated, and monitor leftover risk after launch.
• Create AI-assisted pipelines that monitor research and sort incoming signals so analysts focus on judgment calls.

What it asks for

• Ability to judge technical AI research and evaluation results and explain them clearly to non-specialists (required).
• English proficiency (required).
• Familiarity with key AI labs and evaluators (preferred).
• AI evaluation experience across modalities and Python or SQL scripting for analysis (preferred).
• Knowledge of misalignment research such as deception, evaluation awareness, and reward hacking (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.

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