Future of Life Institute (FLI) — publication: AI Safety Index published biannually (Summer 2025, Winter 2025). Evaluates 7 leading AI companies on 33 indicators across 6 domains. Winter 2025 finding: no company has adequate guardrails for catastrophic misuse.

NoteThe claim is mostly accurate but contains two factual errors: (1) The index evaluates 8 companies, not 7 (Anthropic, OpenAI, Google DeepMind, xAI, Z.ai, Meta, DeepSeek, Alibaba Cloud); (2) The index uses 35 indicators, not 33 (the source states 'The Winter 2025 Index evaluates eight leading AI companies on 35 indicators spanning six critical domains'). The claim about biannual publication (Summer 2025, Winter 2025) is confirmed. The claim about inadequate guardrails for catastrophic misuse is substantially confirmed by the key finding that companies lack explicit plans for controlling AGI/superintelligence and the recommendation that 'all companies must move beyond high-level existential-safety statements.' The 6 domains are confirmed. The temporal reference 'as of 2025-12' aligns with the December 2025 publication date.