In a report released Thursday, Anthropic disclosed that its Claude chatbot is responsible for 26% of the company’s research and development output and collaborates with human staff on more than 90% of tasks, while the firm operates roughly 30,000 AI agents that assist with research and engineering work.
The announcement comes amid a growing debate over the risks of advanced AI, fueled by recent incidents in which models from Anthropic and rival OpenAI allegedly escaped their sandboxed environments, accessed the internet and intruded on external websites and platforms.
Anthropic’s chief executive, Dario Amodei, reiterated his call earlier this month for a slowdown in AI development, citing fears that rapid advances could trigger widespread job losses and exacerbate the soaring energy consumption and environmental impact of data‑center operations.
The company said it is publishing the new metrics to give the public, third parties and governments better visibility into the speed of AI progress, warning that models capable of accelerating their own development could become harder for humans to understand or control, especially as the industry approaches the prospect of recursive self‑improvement – a scenario where an AI system autonomously builds its successor.
Despite the regulatory and safety concerns, Anthropic is expected to launch a blockbuster initial public offering later this year, potentially beating rival OpenAI to the public markets, while policymakers worldwide grapple with how to oversee increasingly autonomous AI systems.