Anthropic says Claude is helping build future AI models, reveals how much work it handles
Follow UsAnthropic has revealed that its Claude AI models are playing an important role in the development of future AI systems.
The company has also shared how much of its artificial intelligence research and development (R&D) work is handled by Claude
As of August 2026, Claude leads 26 per cent of Anthropic’s AI R&D work
Anthropic has revealed that its Claude AI models are playing an important role in the development of future AI systems. The company has also shared how much of its artificial intelligence research and development (R&D) work is handled by Claude. As of August 2026, Claude is able to lead 26 per cent of Anthropic’s AI R&D work, while more than 90 percent of the work involves AI at some level. However, Anthropic says Claude is not yet fully autonomous in any of the areas it measures. The company says these metrics are meant to show how close AI is getting to developing its successors with less human involvement.
SurveyAnthropic calls the new measurement the Anthropic R&D Automation Index. The index tracks different types of AI research and development work at the company. Each task is given an automation rating based on how much AI can handle.
Also read: Apple iPhone 18 Pro Max and iPhone 18 Pro sale now live: Price, bank discounts and more details
AI systems are getting more powerful, and they're increasingly being used to build the next version of themselves. We want to illuminate that progress for the public.
— Anthropic (@AnthropicAI) September 17, 2026
Today, we're sharing three measurements that help track AI development:
1. How much AI R&D is done by AI.
2. How…
The scale ranges from AL0, where there is no AI involvement, to AL5, where AI can complete work fully on its own without a human in the loop. Anthropic says Claude currently reaches the AL4 level, described as “leads.” At this level, AI can handle most of a task from a high-level prompt, while a human supervises the process.
Anthropic also shares new data on how it monitors AI agents working inside the company. “As of August 2026, there were approximately 30,000 agents doing research and engineering work at Anthropic at any one time in our most-used internal platform,” the company revealed.
The company measures how often agent actions are checked, how quickly those actions are reviewed and how often they are blocked, redirected or flagged for further review.
Anthropic says individual agents “rarely misbehave” in its monitoring data so far. However, it points out that even rare problems could become more frequent if AI agents eventually operate at a much larger scale.
The company says these measurements are important as AI systems become more capable and increasingly used to build future AI models.
“It is therefore important to share these metrics to understand how close the world is to reaching recursive self improvement,” Anthropic said.
