No major AI developer earned better than a C plus in the latest AI Safety Index, according to a Visual Capitalist analysis of scoring published by the Future of Life Institute. The index grades companies across six categories: risk assessment, current harms, safety frameworks, existential safety, governance and accountability, and information sharing.

Anthropic, the developer of Claude, took the highest overall grade at C plus. It was the only company to earn an A range mark in two categories, with an A minus in governance and accountability and an A minus in information sharing. OpenAI followed with an overall C, scoring higher than Anthropic in the current harms category, helped by the fact that it was the only company with a published whistleblowing policy at the time of the report. Google DeepMind received a C minus overall. xAI and Meta both landed at D. Zhipu AI and DeepSeek each received a failing overall grade of F.

Existential safety produced the weakest results across the board, with no company scoring above a D and five of the seven companies receiving an F. Only three companies, Anthropic, OpenAI and Google DeepMind, reported any testing for high risk capabilities such as biological or cyber attack misuse, and the index found that even those efforts often lacked clear reasoning or rigorous standards.

The Future of Life Institute concluded that AI capabilities are advancing faster than the risk management practices meant to contain them. The index also noted that stricter national AI rules in China may explain the weaker showing by Zhipu AI and DeepSeek on self governance and transparency measures.

Source: Visual Capitalist - https://www.visualcapitalist.com/visualized-ai-safety-report-card-of-leading-companies/