Anthropic says AI may soon accelerate research

Anthropic said internal estimates and expectations about future models indicate it may imminently meet its AI-accelerated progress threshold.[2] It also said it has not designated any current model as exceeding that threshold.[2] The Responsible Scaling Policy describes one case involving twice the rate of progress in overall AI capability.[3] It sets a separate threshold for models that could substitute for Anthropic’s full research staff at a cost no more than five times as high.[3] Anthropic’s policy says the automated-research-and-development threshold is intended to capture the onset of dramatic recursive self-improvement.[3]

Anthropic’s August 2026 report said its internal research and development was significantly faster with AI assistance, but not yet twice as fast.[4] The company also said Claude writes a large majority of the code merged into its production codebases.[4] Anthropic warned that using AI to accelerate research and development could allow model capabilities to advance faster than safety measures.[2] It has previously failed to identify a significant acceleration in its development rate until after it occurred.[2] The assessment concluded overall progress remains below the RSP threshold, though that conclusion is uncertain and subject to lag.[3]

Anthropic said it does not expect to safely let Claude build successive generations of frontier models autonomously without meaningful human supervision or understanding.[2] It is considering updates to its Responsible Scaling Policy to keep humans in control of frontier-model development.[2] One option under consideration is requiring human review and approval of some research and development decisions as models do more of that work.[3] For teams evaluating AI-assisted development, Anthropic describes increased speed as significant, while saying it does not expect safe autonomous frontier-model development without meaningful human involvement.[4][2]

A filing was filed with the Securities and Exchange Commission on November 2, 2026.[1] It says Anthropic applied to list its Class A shares on the Nasdaq Global Select Market under the ticker ANT.[1] The proposed IPO price per share remained blank in the amended filing.[5] Anthropic reported about $895.4 billion in purchase and lease commitments, largely for compute and infrastructure.[5] These disclosures provide information about the planned listing and the company’s commitments, but the cited material does not give a proposed per-share price.[5][1]