Agentic Cyber Explorerby Fide AI · PreviewPreview
Chronicle/Policy & standards

DeepMind Frontier Safety Framework v3.1 raises security for its cyber critical capability level

PolicyFrameworkSignificance assistant-drafted

Google DeepMind's Frontier Safety Framework version 3.1 introduced Tracked Capability Levels for earlier warning in CBRN and ML R&D and misalignment, and raised the recommended security for the CBRN, cyber and harmful manipulation CCLs to Security Level 2+. The cyber CCL definition itself, Cyber uplift level 1, is unchanged from v3.0.

Why it matters

It raises the recommended weight security for cyber-capable models to Security Level 2+, adding protection against non-state and insider theft, while keeping the 2025 rationale that higher security levels are likely not warranted.

Key facts

As stated in the sources, with where to find them.

  • Version 3.1 (April 17, 2026) raised security for CBRN, Cyber and Harmful Manipulation CCLs to Security Level 2+ to protect against non-state actors and insider threats.Section 5.3 Past Updates and Changes
  • In v3.1, Table 2.2.2.a pairs Cyber uplift level 1 with Security level 2+ (v3.0 said Security level 2).Section 2.2.2, Table 2.2.2.a
  • TCLs were introduced for CBRN and for ML R&D and misalignment, not for cyber.Section 1.2

Findings that cite this record

No tracked finding cites this record yet.

Sources

Related records