AI Safety and Alignment Researcher

AI Breaking Wire; Greater London; Hybrid; GBP 90,000 - 150,000; Full time

Benefits offered by this job

Equity / stock options

Health, wellness & family care

World-class computing resources

Generous sabbatical and time-off

Job summary

Google DeepMind is seeking an AI Safety and Alignment Researcher to develop rigorous empirical and theoretical frameworks that keep advanced AI systems safe, robust, and aligned with human values.

The role focuses on foundational and applied research across model interpretability, robustness, and scalable oversight, with strong ties to product and engineering teams to embed safety into deployments.

Qualifications

• PhD (preferred) in a quantitative field with strong research track record.
• Strong background in ML safety, alignment, interpretability, or adversarial robustness.
• Proficiency in Python and deep learning frameworks (TensorFlow or PyTorch).
• Excellent communication skills with peer-reviewed publications.

Responsibilities

• Conduct fundamental and applied research into model interpretability, robustness, and scalable oversight.
• Design evaluation benchmarks to test model vulnerabilities against jailbreaks, prompt injection, and deceptive alignment.
• Partner with product and engineering teams to integrate safety guardrails into deployment pipelines.
• Collaborate with academic institutions and external research bodies on safety standards.

Skills

ML safety

Interpretability

Adversarial robustness

Academic publishing

Education

PhD in CS/Math/Physics/Philosophy

Tools

TensorFlow

PyTorch

Job description

About the Role

Google DeepMind is at the forefront of artificial intelligence research. We are seeking an AI Safety and Alignment Researcher to develop empirical and theoretical frameworks for ensuring advanced AI systems remain safe, robust, and aligned with human values.

Responsibilities

• Conduct fundamental and applied research into model interpretability, robustness, and scalable oversight.
• Design rigorous evaluation benchmarks to test model vulnerabilities against jailbreaks, prompt injection, and deceptive alignment.
• Partner with product and engineering teams to integrate safety guardrails into deployment pipelines.
• Collaborate with academic institutions and external research bodies on safety standards.

Requirements

• Advanced degree (Ph.D. preferred) in Computer Science, Mathematics, Physics, or Philosophy with a heavy quantitative focus.
• Strong background in machine learning safety, alignment, interpretability, or adversarial robustness.
• Proficiency in Python and deep learning frameworks (TensorFlow or PyTorch).
• Excellent communication skills with a proven track record of peer-reviewed publications.

Benefits

• Industry-leading compensation including base salary and equity/stock options.
• Comprehensive health, wellness, and family care benefits.
• Access to world-class computing infrastructure and research resources.
• Generous sabbatical and time-off policies.