A report by the Future of Life Institute, an AI safety research institute based in the US, recently announced a safety ranking with 9 leading AI companies in the world.
To implement this report, 7 researchers and management experts identified the ranking based on public data and information provided by companies. The group assessed efforts in 6 different categories: Risk assessment, current harm, safety framework, vital safety, governance and accountability, information sharing.
No AI company received an "A" in any category, but the US artificial intelligence company - Anthropic achieved the overall best score of "C+". Meta rose 2 places to 4th place, while xAI dropped 3 places to 7th place, right before DeepSeek of China and Mistral of France, the last 2 companies.
Mistral was first included in the list, but when asked by AFP about the final position, the company said that the framework of the report was not consistent with the company's AI model development method. The French company develops open models, allowing users to download and edit products.
Many of Mistral's competitors that develop closed AI models, including Anthropic, OpenAI and Google DeepMind, are also included in the list.
I am very disappointed to see them last, especially because Europe has really led in AI safety" - Max Tegmark, MIT Professor and President of the Future of Life Institute, told AFP. "We have contacted many times" but Mistral did not answer the organization's survey, he added. Alibaba, xAI and DeepSeek also did not answer the survey.
3 Chinese AI developers included in the report also produce open-source models and are in the bottom half of the ranking: DeepSeek (5th), Alibaba Cloud (6th) and Z.ai (8th).
The Future of Life Institute report noted that some companies that previously prohibited the use of their technology in the military have "gradually changed their views", including Anthropic.
According to the report, all 9 companies failed to counter "existing" threats such as pursuing models that reach human-level intelligence, called "general artificial intelligence" or AGI. Despite "constructive efforts", widespread efforts are "completely insufficient".
Other risks include the possibility of abusing the model to carry out a cyberattack or perform tasks that are potentially harmful to humans.
Anthropic recently attracted attention when releasing the most powerful AI model of Mythos. In early April, the company based in San Francisco, USA released Mythos to a few trusted organizations due to its ability to detect cybersecurity vulnerabilities for bad actors.
