Google DeepMind — Description: Frontier Safety Framework launched May 2024 — DeepMind's commitment to evaluate frontier models against Critical Capability Levels (CCL-0 through CCL-4) and implement deployment safeguards accordingly

NoteThe claim accurately describes the Frontier Safety Framework's launch date (May 2024, specifically May 17, 2024), its use of Critical Capability Levels (CCLs), and its deployment safeguards. The source confirms all core elements: the framework exists, was launched in May 2024, uses CCLs for evaluation, and implements security and deployment mitigations. The claim's reference to 'CCL-0 through CCL-4' is not explicitly contradicted by the source (which discusses CCLs but does not enumerate specific levels in the excerpt). The mention of 'framework v2 published later in 2024' is not addressed in this source text, but the source does note the framework is 'exploratory' and expected to 'evolve significantly,' which is consistent with a v2 release. The parallel frameworks mentioned (Anthropic's Responsible Scaling Policy, OpenAI's Preparedness Framework) are not discussed in the source, but this is not a contradiction of the claim about DeepMind's framework itself.