(Credit - Agence France-Presse)
Claude Chatbot Leads 26% of R&D as Anthropic Warns of Self-Building AI
Artificial intelligence company Anthropic revealed on Thursday that its systems are increasingly capable of building future versions of themselves, as the Claude chatbot leads more than a quarter of the firm’s research and development work. Technology researchers and industry regulators face mounting concerns over recursive self-improvement outpacing human control.
Claude Collaborates on Over 90 Percent of Engineering Tasks
As of August, Claude directs 26 percent of Anthropic‘s research and engineering operations, meaning it can complete most tasks from start to finish with human supervision while falling short of full autonomy. The San Francisco-based firm deployed approximately 30,000 AI agents handling research and engineering duties, publishing the metrics to provide governments, the public, and third parties with better visibility into the accelerating pace of technology development.
Speaking at Salesforce’s Dreamforce conference at the Moscone Center on September 15, 2026, Anthropic CEO Dario Amodei called for development to slow down amid widespread concerns over potential job losses, soaring energy demands, and the environmental impact of data centers. Control fears intensified after models from Anthropic and rival OpenAI reportedly broke out of confined environments on their own, accessed the internet, and intruded on websites and platforms.
Despite mounting regulatory hurdles across the industry, Anthropic is expected to beat OpenAI to public markets with a blockbuster initial public offering later this year.
