New AI Safety Rankings Reveal Stagnant Progress, No Company Earns Top Marks

The latest AI Safety Index has assigned middling grades to leading artificial intelligence companies, including Anthropic, OpenAI, Google DeepMind, and Meta, with little evidence suggesting the industry is becoming safer overall. Experts note that even the highest-scoring companies achieved only mediocre marks, and that industry competition alone is not fostering greater safety.

One notable improvement came from Meta, which climbed from a D to a D+ grade. "It's encouraging to me that a company can improve so much in just six months," said Max Tegmark, co-founder and president of the institute that compiled the rankings. However, this progress did not extend to Elon Musk's xAI, which rebranded as SpaceXAI following a merger with his other ventures. The company fell to an F grade, joining DeepSeek from China and Mistral from France among the lowest scorers. Tegmark highlighted this distribution, stating, "The worst scorers come from three different continents, shows this is a global problem."

Tegmark suggested that a true "race to the top" in AI safety will necessitate regulatory intervention. He expressed cautious optimism, pointing to developing regulations such as the European Union's AI Act, Chinese rules set to take effect soon, and a more risk-aware approach from the U.S. administration. "We're rapidly going towards a place where there's a global agreement about at least basic safety standards," he commented, indicating a potential for international alignment on crucial safety benchmarks.