Google DeepMind recently released "An Approach to Technical AGI Safety and Security", outlining how it thinks about severe risks from advanced AI systems. The paper focuses on two central concerns: misuse, where humans use AI systems for harmful purposes, and misalignment, where AI systems pursue behavior their developers did not intend.

What does the document reflect about the safety mindset at Google DeepMind? Is there substantial difference to Anthropic or OpenAI’s present and past published frameworks? How much AI safety research actually happens at Google DeepMind these days?

Does any of this still matter, in a time of increasing political control over frontier models?

Artificial General Intelligence (AGI) promises transformative benefits but also presents significant risks. We develop an approach to address the risk of harms consequential enough to significantly harm humanity. We identify four areas of risk: misuse, misalignment, mistakes, and structural risks. Of these, we focus on technical approaches to misuse and misalignment.

For misuse, our strategy aims to prevent threat actors from accessing dangerous capabilities, by proactively identifying dangerous capabilities, and implementing robust security, access restrictions, monitoring, and model safety mitigations.

To address misalignment, we outline two lines of defense. First, model-level mitigations such as amplified oversight and robust training can help to build an aligned model. Second, system-level security measures such as monitoring and access control can mitigate harm even if the model is misaligned. Techniques from interpretability, uncertainty estimation, and safer design patterns can enhance the effectiveness of these mitigations. Finally, we briefly outline how these ingredients could be combined to produce safety cases for AGI systems.

— Rohin Shah et al., An Approach to Technical AGI Safety and Security (2026)
— Robert Wiblin, Rohin Shah in 80,000 Hours Podcast: What it’s really like to run AGI safety at Google DeepMind (2026)