AI Safety Index: No lab scores above C+
What it really says
The Future of Life Institute published its second annual AI Safety Index, evaluating nine leading AI labs across six domains: Risk Assessment, Current Harms, Safety Frameworks, Existential Safety, Governance and Accountability, and Transparency and Communication. Anthropic received the highest grade of C+ and leads in five of six domains through comparatively strong transparency, an established safety framework, and technical research. OpenAI dropped from C+ to C but leads in Risk Assessment. Google DeepMind also received a C. Meta improved from sixth to fourth place with a D+. Three labs received an F: xAI from Elon Musk, which dropped from fourth to seventh place, China's DeepSeek, and Europe's Mistral. Particularly concerning: Anthropic, OpenAI, Google DeepMind, and Meta had publicly pledged to pause development if certain risk thresholds were reached. According to the report, all four have since weakened or effectively abandoned these commitments. Additionally, several companies that previously excluded military AI applications have reversed this position.
Our assessment
Nine AI labs worldwide score C+ at best, with three failing entirely. The result shows that voluntary industry safety commitments carry little weight: all leading companies have weakened their promises to pause development when risks emerge. On the positive side, the index itself creates public accountability pressure on labs. The assessment validates the European regulatory approach, which relies on binding rules rather than self-commitments. Particularly notable: Mistral, Europe's only frontier AI lab, receives an F, raising questions about whether Europe's own AI champion takes seriously the rules the EU is currently enforcing. The reversal on military AI applications deserves particular attention as it signals a fundamental values shift in the industry.
Relevance for Germany
The report has several implications for Germany. First, it confirms the necessity of the EU AI Act and Germany's recently passed AI Implementation Act: voluntary industry safety commitments are systematically being eroded. Second, Mistral's F grade is remarkable because the French company is considered Europe's answer to OpenAI and Google. If Europe's champion fails on safety, the question arises whether EU regulation is sufficient or enforcement too weak. Third, German companies heavily use AI models from Anthropic, OpenAI, and Google, and the index gives them guidance on which provider to trust more on safety matters. Fourth, the reversal on military AI is relevant to Germany's security policy debate, particularly given discussions about AI in the Bundeswehr.
Fact check
The AI Safety Index Summer 2026 was published on July 1, 2026 by the Future of Life Institute. The grades are consistently reported by Time, IEEE Spectrum, Seeking Alpha, and MIT Sloan Management Review: Anthropic C+, OpenAI C, Google DeepMind C, Meta D+, xAI F, DeepSeek F, Mistral F. The methodology is based on evaluation by independent experts across six domains. The claim that safety commitments have been weakened comes directly from the report and is confirmed by Axios and Seoul Economic Daily. The reversal on military AI is also documented in the report. OpenAI's drop from C+ to C is confirmed by multiple sources.
Source
- • https://futureoflife.org/ai-safety-index-summer-2026/
- • https://futureoflife.org/wp-content/uploads/2026/07/AI-Safety-Index-Report_010726_2Pager.pdf
- • https://time.com/article/2026/07/07/ai-safety-rankings-openai-anthropic-meta/
- • https://seekingalpha.com/news/4611964-google-anthropic-openai-lead-ai-safety-index-spacexai-receives-f