Google DeepMind Opens Multiple AGI Safety and Alignment Research Positions for 2026
Overview
Google DeepMind has announced multiple openings in its AGI Safety and Alignment Team (ASAT), announced in July 2026. The team, led by Rohin Shah, works to reduce existential risks from highly capable AI by developing technical solutions that can guide the behaviour of future advanced systems. Its work covers aligning advanced AI with human goals, safeguards against misaligned deployments, coordinated approaches to safety and governance, stress testing of frontier models and deep alignment research. Rather than hiring for one profile, the team is looking for exceptional researchers and engineers across several disciplines. Most hires are expected to be in London, United Kingdom, or San Francisco, United States, with flexibility on location where appropriate.
The team aims to produce technical evidence that resolves scientific uncertainty rather than focusing on advocacy, to build foundations for future AI governance, to prioritise research relevant to future generations of highly capable models and to keep a coherent long-term strategy. Future directions include deep alignment, AI control, research on systems whose chain-of-thought is no longer directly observable, model forensics and interpretability, blue-team defences, red-teaming and stress testing, and amplified oversight. Staff work with frontier models and large-scale computing, collaborate with leading researchers and can advise external research projects, and many former members have moved to influential roles in AI organisations and public-sector AI safety institutes.
Who can apply
The team seeks candidates who share its mission, are interested in long-term challenges of advanced AI, can work independently in uncertain and fast-moving research settings, collaborate well and have strong software engineering skills. Experienced researchers are encouraged, though some roles stress engineering over academic research experience. Candidates should be comfortable with AI-assisted coding tools, as engineering interviews allow coding agents.
How to apply
Applications go through separate portals for the United States and the United Kingdom. Applicants are encouraged to add a concise cover letter explaining why they fit the team and why they want to contribute to Google DeepMind's safety mission. The process includes CV screening, a short screening interview on technical ability and mission alignment, about four remote technical interviews and team-matching interviews for shortlisted candidates, and typically takes several months.