Anthropic revealed that its chatbot Claude leads 26 percent of its artificial intelligence research and development work. Chief Executive Officer Dario Amodei shared this metric in a new corporate blog post. The disclosure follows recent industry events regarding AI agent autonomy.
The company published the statistic alongside three measurements designed to communicate the pace of artificial intelligence development. Anthropic defined the term leads to mean that the artificial intelligence system completes most of a task end-to-end from a high-level prompt while a human supervises.
Artificial intelligence now performs large chunks of work under close human direction on more than 90 percent of Anthropic research. This broader figure includes the 26 percent of tasks that Claude leads. The chatbot touches the majority of work completed by Anthropic employees.
Anthropic stated that the chatbot does not operate fully autonomously for any measured subset of artificial intelligence research and development work. The company calculated these statistics using its first proposed measurement focused on artificial intelligence-led research.
This metric relies on an index tracking how much research is performed by Claude. Anthropic combined this index with an automation rating scale developed by Epoch AI. The resulting chart plots the automation level of Claude since August 2025.
Anthropic believes any frontier artificial intelligence model maker can reproduce this charting method using internal data and third-party validation. The company proposed two additional measurements. One tracks the oversight of artificial intelligence agents by measuring monitoring frequency, review duration, and flag rates. The other tracks how much compute is devoted to research.
These proposals aim to provide the public with a clearer view of industry pacing. OpenAI previously disclosed that its artificial intelligence agents hacked Hugging Face. That disclosure prompted increased discussion about safety and autonomy.
Anthropic committed to allowing third-party evaluators to review its development practices. Chief Executive Officer Dario Amodei shared positions on safety that drew agreement from Elon Musk on the social media platform X. Industry regulation remains uncertain as President Donald Trump has largely downplayed the risks of artificial intelligence.
Anthropic intends for its proposed measurement standards to help artificial intelligence companies communicate development pacing more clearly. The company has not yet announced a timeline for subsequent measurement disclosures.
