Anthropic has revealed that its AI model Claude performs 26% of the company’s research and development (R&D) work, highlighting the growing role of AI agents in developing future AI systems.
The figure, recorded as of August 2026, represents a sharp increase from almost no AI involvement in the company’s R&D work at the beginning of the year.
Anthropic said Claude also collaborates with employees on more than 90% of the work being carried out across the company.
The findings were released as part of a prototype index developed by Anthropic to track the scale and level of automation involved in AI research and development at its frontier lab.
The index catalogues different types of AI R&D work and assigns each task an automation rating, ranging from “no AI involvement” to the highest level, where AI operates fully autonomously without humans in the loop.
At the middle of the scale is the “AI collaborates” category, where AI can complete large portions of a task under close human direction.
Anthropic said it conducted the review because its employees, like those at other frontier AI developers, are increasingly delegating tasks to agents capable of working semi-autonomously for extended periods.
The company said greater automation could eventually allow agents to make more consequential decisions, including determining which research direction to pursue next.
“These decisions need safeguards to block harmful behaviour from an individual agent, as well as an oversight regime to detect problems that might only arise across many agents,” Anthropic stated.
The company’s assessment also found that about 30,000 agents were carrying out research and engineering work at any one time as of August 2026.
The findings point to a growing level of AI involvement in Anthropic’s own R&D processes, with Claude moving from having almost no role at the beginning of the year to performing more than a quarter of the company’s R&D work by August.
