.png){kind=logo}
Anthropic shared new data showing that its Claude AI models now take the lead on about 26% of the company's internal AI research and development work. That number was below 1% just six months ago, in February 2026.
The figures come from a new internal tool Anthropic calls the R&D Automation Index. It tracks how much AI contributes across different research tasks, using a scale built by Epoch AI. The scale runs from AL0, meaning no AI involvement at all, to AL5, meaning a system works fully on its own without any human input.
Claude currently operates at the AL4 level for 26% of the work. At this level, the AI takes on most of a task after receiving a high-level instruction, while a human stays in the loop to supervise. Anthropic was clear that Claude is not running anything fully on its own yet.
The company also noted that more than 90% of its measured AI research now happens at or above the level where AI and humans work side by side. This covers areas like model training, reinforcement learning, evaluation, and engineering.
The numbers behind this shift are significant. In August, around 30,000 AI agents were active at any given time on Anthropic's most-used internal platform. Over one billion agent decisions were tracked that month. Of those, roughly 0.002% were stopped by the company's monitoring system.
Anthropic also shared that about 6% of AI research compute went toward safety work during a week in July. For compute used specifically in AI-driven research, that figure was closer to 12%.
Also Read: Anthropic Eyes USD 2 Trillion IPO as Claude Revenue Surges Past USD 11.5 Billion
Anthropic plans to publish these measurements regularly. The company is also calling on other leading AI developers to release similar data. It believes that shared methods and outside verification could help the broader industry keep pace with how fast AI development itself is changing.
The report marks one of the more detailed looks any major AI company has offered into how much its own tools are shaping the work of building the next generation of AI.
