Anthropic S-1A filing warns of imminent recursive self-improvement threshold

Details

Anthropic's revised S-1A registration statement filed with the SEC on November 2, 2026, discloses that internal estimates indicate the company may imminently meet its Responsible Scaling Policy definition of recursive self-improvement (RSI).
The filing clarifies that none of Anthropic's current models have crossed the RSI threshold, with the company's May 2026 essay confirming that recursive self-improvement is not inevitable and that the recursive loop is not yet closed.
As of August 2026, Claude leads 26% of AI research tasks end-to-end with human oversight, up from under 1% in February, but the share of tasks completed fully autonomously remains zero in every measured area.
Independent METR benchmarks show Anthropic's Mythos Preview reached a 17-hour time-horizon metric in April 2026, doubling roughly every four months since 2023, though wide confidence intervals and measurement ceilings limit the precision of these gains.
The disclosure coincides with Anthropic's move toward a Nasdaq listing under the symbol ANT, with the company noting it has historically not identified significant development acceleration until after it occurred.

Key point