Anthropic says self-improving AI may arrive soon
In an amended S-1/A filing, Anthropic says its internal estimates suggest it may soon meet its own definition of dramatically accelerated progress attributable to AI, the stage known as RSI.
Anthropic, the company behind Claude, added a new warning to its amended S-1/A filing published on November 2, 2026 ahead of its Nasdaq IPO. In the Risk Factors section, the company writes that its internal estimates and expectations for future models indicate it may "imminently" meet its Responsible Scaling Policy definition of "dramatically accelerated progress attributable to AI." That is the company's own term for what the industry calls recursive self-improvement, or RSI.
1. What RSI means
RSI is the point at which AI becomes so good at training new models that its capabilities start compounding. Better models build better models, with less and less human help.
It has not happened yet. Anthropic states in the same filing that none of its current models has been designated as exceeding this threshold. What changed is the expectation: internal estimates about future models suggest the threshold may be reached soon.
2. What the filing actually says
Two admissions stand out in this amendment. First, Anthropic says it already uses AI to help develop new models, and that in the past it failed to identify significant jumps in capability until after they happened.
Second, the company writes that using AI to build more capable AI "could also pose a considerable safety risk, particularly if it sharply accelerates the pace of progress." According to Gizmodo, the Responsible Scaling Policy threshold is when the rate of improvement doubles the baseline rate, with that doubling plausibly attributable to AI.
3. The human oversight promise
Anthropic also made commitments. The company says it does not anticipate being able to safely use Claude to fully autonomously design and build successive generations of frontier models without meaningful human supervision or understanding, even if that becomes technically possible.
It is considering updates to its Responsible Scaling Policy to keep humans in control of frontier model development. WIRED notes the IPO document runs 94 pages of risk factors and uses the phrase "existential risks" three times.
4. Why this sits in an IPO filing
TechCrunch, which first spotted the change, points out a real tension here. Risk Factors are mandatory disclosure for investors, but at the same time they signal to the market how seriously the company takes its own models' potential.
AI Weekly notes the filing lands during a tense IPO window: early buyers are reportedly pushing the valuation toward 1.5 trillion dollars, while an op-ed has urged the SEC to block the deal over the company's compute debts.
5. What to watch
For readers, two things to follow: whether the Responsible Scaling Policy actually gets updated as promised, and whether any Claude model is ever formally designated as crossing this threshold. Anthropic says it is still assessing the underlying data, so no timeline has been given.
What it means for you
This warning comes from the company itself, not outside critics, and it entered a formal SEC document. For everyday users, it is a signal that the people building frontier AI are preparing for a faster rate of progress.
- Expect faster updates: if the pace really accelerates, the AI tools you use today will change more often than they did last year.
- Read the claims carefully: IPO documents have marketing incentives, so "imminently" from an AI company is a signal, not a confirmed date.
- Watch the oversight promise: Anthropic's pledge to keep humans in the loop is a measurable commitment you can check later.
- For builders: if you build on Claude through AI Tools or the API, prepare your workflows for more frequent model changes.