Claude AI Now Leads 26% of Anthropic’s R&D, Company Reports

Anthropic announced that its Claude artificial intelligence model is now leading 26% of the company’s research and development efforts — a figure that stood at zero percent in February. By August, the model was directing one-quarter of all AI R&D activity within the firm.

AI-Driven Development Metrics

The company stated that AI systems are growing more powerful and are increasingly being used to build their own successors. In a blog post, Anthropic said it aims to “illuminate that progress for the public” and outlined three specific measurements to track AI advancement: how much AI R&D is conducted by AI, how…

According to Anthropic, approximately 90% of its research and development work is now carried out in collaboration with Claude. The model handles large portions of the technical workload under precise human supervision, beginning from high-level instructions. Human oversight remains continuous and mandatory throughout the process.

Self-Improvement and Oversight Challenges

Anthropic noted that AI models accelerating their own development may make these systems “harder for humans to understand or control.” The company views transparency around such metrics as critical to gauging how close AI labs are to achieving “recursive self-improvement” — defined as a model’s ability to design and build its own successor.

The firm urged other companies to regularly publish similar metrics. It also reported that roughly 30,000 AI agents were engaged in research and engineering tasks as of August, each operating under monitoring protocols designed to detect inappropriate behavior.

External Safety Oversight and Industry Debate

Anthropic recently committed to appointing third-party external auditors to oversee its safety initiatives. This step follows wider industry discussion on AI safety, intensified after a researcher resigned from Anthropic last week, warning of existential risks posed by the technology to humanity.

Dario Amodei, Anthropic’s CEO, has joined Sam Altman of OpenAI and Elon Musk in calling for a slowdown in AI development. That position has been opposed by other technology leaders and by former U.S. President Donald Trump.