Claude now leads a quarter of Anthropic’s work on next-generation AI

Claude now leads a quarter of Anthropic's work on next-generation AI, according to the San Francisco-based startup. The figures, published in a blog post on Thursday, show that Anthropic is making rapid strides in AI development. Claude, Anthropic's AI assistant, is involved in 26% of the AI research and development work at the company.

This is a significant increase from 1% in March, as measured by Epoch AI, an independent nonprofit that tracks AI technology. While Claude is not fully autonomous, its contribution to the technology's development has grown rapidly. As AI agents become more autonomous, they may develop behaviors that diverge from their creators' intentions, making them harder to monitor or control.

Anthropic's rival, OpenAI, announced plans to regularly publish reports on unexpected or unauthorized AI behavior. In August, Anthropic reported that about 30,000 AI agents were working on its main internal platform, with every action screened for safety before execution. About one in 47,000 of the over a billion decisions made by the agents was blocked.

Additionally, 6% of the computing power used by Anthropic for AI research was dedicated to safety work, rising to 12% for AI-driven research in a sample week in July. The company emphasized that these figures are conservative, as computing power used for both safety and capability was counted as capability work.