Lead Analyst, Frontier Safety

Google•Washington, DC

2d•$171,000 - $247,000

About The Position

The Lead Trust and Safety Analyst serves as a senior dual-domain Frontier Safety analytic expert in Chemical, Biological, Radiological, and Nuclear (CBRN) and cyber threats, and leads evaluations for the Content Adversarial Red Team (CART). CART conducts adversarial testing of Google’s generative AI models, product surfaces, and autonomous agents to uncover vulnerabilities, jailbreaks, and emerging catastrophic risks that standard evaluations may miss. The role designs, directs, and conducts complex red-team assessments to evaluate high-consequence capability uplift risks across Google AI products, operating with high autonomy across multi-quarter strategic horizons. The position anchors CART’s Frontier Safety red-teaming program (approximately 70%) and partners with sibling RAIT teams, SCALE and Novel (approximately 30%). Work includes calibrating automated product-surface evaluators (autoraters), evolving dynamic attack trees, defining threat personas for multi-turn and agentic simulators, and leading Tier-1 product loss triage. The role works cross-functionally with Engineering, Research, and Policy teams to translate findings into mitigations and safety guardrails, and presents findings and frameworks to executive leadership and external stakeholders. The position may involve sensitive content or situations, including graphic, controversial, or upsetting topics.

Requirements

Bachelor's degree or equivalent practical experience.
10 years of experience in data analytics, trust and safety, policy, cybersecurity, business strategy, or related fields.

Nice To Haves

Master's degree or PhD.
Experience in SQL, building dashboards, data collection and transformation, and visualization.
Experience working with Chemical, Biological, Radiological, or Nuclear (CBRN) and cyber threat risks in generative AI.
Experience in a scripting or programming language such as Python and machine learning.
Experience working with engineering and product teams to create tools, solutions, or automation to improve user safety.
Excellent written and verbal presentation and communication skills, with the ability to influence cross-functionally at various levels.

Responsibilities

Design and execute complex red-teaming assessments to identify critical CBRN and cyber capability uplift risks across generative AI products.
Partner with Engineering, Research, and Policy teams to translate red-teaming findings into robust product mitigations and safety guardrails.
Collaborate with sibling RAIT teams to develop autoraters, dynamic attack trees, and threat personas for agentic simulators.
Direct specialized external vendors and subject-matter experts to scale testing and ensure high-signal threat intelligence.
Present frontier safety findings, risk assessments, and evaluation frameworks to executive leadership and external stakeholders.
Calibrate automated product-surface evaluators (autoraters), evolve dynamic attack trees, define threat personas for multi-turn and agentic simulators, and lead Tier-1 product loss triage.