The growing use of artificial intelligence (AI) in all sectors raises a raft of safety concerns, including bias, privacy violations and cybersecurity threats. How seriously are AI companies managing the risks that come with more capable models?
To gauge this issue, safety scores have been assigned to major AI companies using data from the AI Safety Index published by the Future of Life Institute. The assessment considers six key metrics: risk assessment, current harms, safety frameworks, existential safety, governance and information sharing.
Anthropic, the creators of Claude, earned an overall safety grade of C+ and was awarded an A- in both governance and accountability along with information sharing. OpenAI, creators of ChatGPT, received a C grade overall. The only category in which it scored higher than Anthropic was in the current harms metric, partially thanks to the fact that OpenAI was the only company with a published whistleblowing policy at the time of the report’s publication.
Only three of the companies considered — Anthropic, OpenAI and Google DeepMind — report substantive testing for dangerous capabilities linked to large-scale risks such as bio- or cyber-terrorism.
Chinese companies Zhipu.AI and DeepSeek both received failing grades overall. China’s stricter national AI regulations may explain their weaker performance on Western-aligned self-governance and transparency metrics.
While AI holds tremendous potential, it also poses significant risks that require careful consideration and proactive management to ensure its safe and ethical use in society.
