Anthropic's amended S-1/A says that "our internal estimates and our expectations regarding future models, rather than our current models, indicate that we may imminently meet our Responsible Scaling Policy's definition of 'dramatically accelerated progress attributable to AI.'" The same passage says: "We have not designated any of our current models as exceeding this threshold, and we continue to assess the underlying data."

What changed

We covered the amended filing's compute and lease commitments and the original filing's headline numbers. This passage is about something else: how fast Anthropic's filing suggests AI may speed up its own research, in our reading.

Read plainly, it is a statement about a definition in the company's Responsible Scaling Policy. Anthropic says its internal estimates point to a possible crossing of one of those thresholds. The estimate is tied to future models, and the filing says no current model has been designated as over the line. The word is "may," not "has."

The filing adds: "We are also evaluating the implications of meeting the AI research and development capability threshold and will release further reports in the future as we set forth in our Responsible Scaling Policy."

What the filing says about AI building AI

The filing describes the mechanism in its own words: "Each version of a Claude model assists in building the next, helping to compress development timelines and produce compounding improvements in the speed at which we innovate."

It backs this with a usage figure: "The value of our own internal Claude usage at list prices has grown by over 250 times from January 2026 to September 2026, and in April 2026, this monthly value first exceeded our monthly personnel expense, excluding stock-based compensation." That is a measure of how much Claude Anthropic uses internally, valued at list prices.

The filing also names the downside: "Using AI to help build more capable AI could also pose a considerable safety risk, particularly if it sharply accelerates the pace of progress."

Safeguards and what is only being considered

Anthropic draws a line at full automation: "However, we do 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, if and when that becomes technically possible."

From that, the filing says: "Accordingly, we are considering updates to our Responsible Scaling Policy to keep humans in control of frontier model development at Anthropic." It adds: "We are also considering additional measures, including updates to our Responsible Scaling Policy, to require human review and approval of certain decisions in our research and development processes as our models perform an increasing share of that work."

The key word is "considering." The filing does not say these updates have been adopted, and the passages quoted here do not say they have been.

The delayed pre-training run

One concrete decision appears in the filing: "we recently chose to delay pre-training a new frontier base model for more than a month to allow time to further enhance our security measures related to preventing model weight exfiltration and protecting our intellectual property." The stated reason is security around model weights and intellectual property. The passage we quote does not tie the delay to the accelerated-progress threshold, and we do not either.

How the coverage framed it

TechCrunch headlined its report "Anthropic warns that it's very close to RSI capability, aka self-building AI". That is the outlet's framing. A search of the filing text found no use of "very close" or the abbreviation "RSI."

The nearest wording is in a passage where the Responsible Scaling Policy "sets out our plans to monitor and mitigate the risks as we approach what has been termed 'recursive self-improvement.'" The filing gives no definition of that term, so we do not supply one. What it does say in its own terms is the "may imminently meet" language above, about a defined policy threshold and future models.

Who is affected

Most directly, prospective investors, who read the filing. Anthropic's customers and developers are affected only indirectly: the passages quoted here announce no product change. In our view, policymakers and safety researchers now have a new company statement to read against the Responsible Scaling Policy.

What this means for you

If you use Claude, the passages quoted here announce no change to the product. They describe an expectation about future models and a set of policy updates under consideration. The company's own wording is cautious, and a headline that says a company is "very close" to a capability goes further than the text. This is not investment advice, and we draw no conclusion about the offering.

What we checked

The quotations are from the S-1/A on EDGAR, as extracted from its text. We did not review the Responsible Scaling Policy itself for this story, so we cannot say what its definition of "dramatically accelerated progress attributable to AI" requires beyond what the filing quotes. We link the outside coverage quoted above for its framing only and rely on it for nothing else.