Anthropic says its growing use of artificial intelligence in research and development could bring a major shift in how quickly new AI capabilities advance. In a company filing, it said internal estimates suggest future models may soon meet its Responsible Scaling Policy's definition of "dramatically accelerated progress attributable to AI."
The company has not designated any current model as having crossed that threshold and says it is still assessing the data. Anthropic also noted that AI tools have helped develop models whose capability gains were not immediately recognized, highlighting the challenge of keeping evaluation and safeguards in step with progress.
Human oversight remains part of the plan
Anthropic uses AI in its development work, while acknowledging that this could make progress faster than its ability to validate safety, alignment and security measures. Its Responsible Scaling Policy addresses how the company will monitor and manage risks as it approaches what is commonly called recursive self-improvement--AI systems helping to improve the development of future AI.
The company said it does not expect to use Claude to design and build successive generations of advanced models entirely autonomously without meaningful human supervision or understanding. It is considering policy updates intended to keep people involved in frontier-model development.
Anthropic's statement describes a possible future threshold, not a claim that current systems can independently build their successors. If AI-assisted research continues to accelerate, careful evaluation and human oversight will be central to ensuring that progress remains useful and responsibly managed.
