Introducing the Frontier Safety Framework
Google DeepMind's Frontier Safety Framework sets out protocols for identifying and mitigating severe risks from advanced AI models. It focuses on four risk domains—autonomy, biosecurity, cybersecurity, and machine learning R&D—by establishing Critical Capability Levels, early warning evaluations, and targeted security and deployment mitigations.
Summary and content fields are drawn from the document text and are pending editorial review.
Who issued it
Who it governs
AI developers, researchers, and policymakers interested in safety practices and risk management frameworks for advanced frontier AI models.
What the document contains
The document covers the components of Google DeepMind's Frontier Safety Framework, including identifying Critical Capability Levels, performing early warning evaluations, and implementing security and deployment mitigations across specific high-risk domains.
The text does not provide specific numerical capability thresholds, exhaustive technical evaluation datasets, or legally binding enforcement mechanisms.
- The Frontier Safety Framework establishes protocols for identifying and mitigating future severe risks from advanced AI models.
- It relies on defining Critical Capability Levels across risk domains such as autonomy, biosecurity, cybersecurity, and machine learning R&D.
- Early warning evaluations will be used to periodically test models and detect when they approach critical capability thresholds.
- Tailored security and deployment mitigations will be applied when models pass early warning evaluation thresholds.
- The framework is exploratory and designed to evolve over time through implementation and collaboration with industry, academia, and government.
· Identifying capabilities a model may have with potential for severe harm · Evaluating our frontier models periodically to detect when they reach these Critical Capability Levels · Applying a mitigation plan when a model passes our early warning evaluations
How it scores
Each score is backed by a quote verified against the document text. Editorial review has not yet confirmed them.
Accountability
1/5Clear responsibility for AI outcomes
We have invested significantly in our Frontier Safety Team, which coordinated the cross-functional effort behind our Framework.
Investing in the science
screening score · 2026-10-23
Fairness
0/5Non-discrimination and equitable treatment
screening score · 2026-10-23
Human Dignity
0/5Respect for the inherent worth of persons
screening score · 2026-10-23
Human Oversight
0/5Humans remain in control of consequential decisions
screening score · 2026-10-23
Privacy
0/5Protection of personal data and private life
screening score · 2026-10-23
Transparency
3/5Openness about how AI systems work and are used
We hope that sharing our approaches will facilitate work with others to agree on standards and best practices for evaluating the safety of future generations of AI models.
Staying true to our AI Principles
screening score · 2026-10-23
Enforcement
4/5Consequences and compliance mechanisms with real force
Applying a mitigation plan when a model passes our early warning evaluations. This should take into account the overall balance of benefits and risks, and the intended deployment contexts.
The framework
screening score · 2026-10-23