Anthropic says its AI may soon be speeding up its own progress

Anthropic, the maker of Claude, filed an amended S-1 with the Securities and Exchange Commission on Nov. 2, and its most striking addition is a sentence rather than a number. The Anthropic amended S-1 says the company's internal estimates indicate it "may imminently meet" the bar its own safety policy sets for "dramatically accelerated progress attributable to AI." The same filing adds third-quarter results, showing nearly $17 billion in quarterly revenue and close to $900 billion in spending commitments, while the offering's price, size and date remain blank.

What the Anthropic amended S-1 says about AI progress

The new language sits in a risk factor about Anthropic using its own models to speed up its research. The filing describes a process it calls "closing the loop," in which "successive versions of our models assist in building the next." It warns that this "may cause the capabilities of our models to advance more rapidly than our ability to develop, validate, and implement corresponding safety, alignment, and security measures."

The original S-1, filed Oct. 26, already said Anthropic "may not identify a significant acceleration in our rate of development until after it has occurred." The amendment rewrites that line to say "we have in the past not identified, and may in the future not identify, a significant acceleration in our rate of development until after it has occurred." It then adds: "For example, 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 filing also says the evidence is imperfect. "Much of our evidence regarding the pace of such AI-driven acceleration is derived from lagging indicators," it states, and the company's ability to measure how fast its capabilities are advancing "is limited."

Has Anthropic crossed its AI progress threshold?

No. The filing says: "We have not designated any of our current models as exceeding this threshold, and we continue to assess the underlying data." The estimate concerns future models, not the ones Anthropic sells today. The company adds that it is "evaluating the implications of meeting the AI research and development capability threshold" and will release further reports as its Responsible Scaling Policy requires.

The reason this belongs in an IPO prospectus is that safety steps can cost money and time. Anthropic says its precautionary measures have included "delaying the release of models and limiting their initial availability," and that future measures "could slow our capability gains relative to competitors that do not adopt similar constraints." A safety threshold is now written down as a business risk that shareholders are being asked to accept.

The third-quarter numbers

The amendment fills in the quarter the original filing lacked. Revenue was $16.96 billion in the three months ended Sept. 30, up from $11.56 billion in the second quarter and $1.38 billion a year earlier. For the nine months, revenue was $33.24 billion, compared with $2.55 billion in the same period of 2025.

Growth is still extraordinary but is slowing from quarter to quarter. CNBC calculated that third-quarter revenue rose around 47% from the second quarter, down from 144% growth in the second quarter and 132% in the first.

Losses remain large. The net loss was $23.29 billion in the third quarter, down from $66.31 billion in the second quarter but up from $12.14 billion a year earlier. Over nine months the net loss reached $104.04 billion, against $25.11 billion a year before; CNBC noted that most of it came from the change in fair value of convertible notes and derivative liabilities. On an operating basis, Anthropic lost $567 million in the third quarter after reporting operating income of $974 million in the second.

Close to $900 billion committed to compute

The amended filing says Anthropic has entered into purchase commitments and lease agreements with an aggregate committed spend of approximately $895.4 billion, "substantially all of which is related to compute and infrastructure."

Spending is already rising fast. Compute and infrastructure expenses were $18.81 billion for the nine months ended Sept. 30, compared with $4.38 billion a year earlier. "As our compute and infrastructure capacity grows, the associated compute and infrastructure expenses have increased," the filing says.

What is still missing

The amendment does not include a price range, a share count or an assumed offering price; those spaces in the prospectus are still blank. There is no valuation in the filing, and Anthropic has not announced a listing date. Earlier press reports pointed to a debut in the second half of November, with IFR reporting Nov. 20, but those timelines are unconfirmed by the company.

Why it matters for people building things

For founders and investors, the filing puts two forces side by side. One is commercial: revenue growing more than tenfold in a year, paid for with hundreds of billions of dollars in compute promises. The other is technical: a company saying in a legal document that AI may soon be meaningfully accelerating its own development, and that it may slow down if safety work does not keep pace.

The next things to watch are a further amendment with a price range, any report Anthropic releases under its Responsible Scaling Policy on the research threshold, and whether quarter-over-quarter growth keeps cooling as the company heads into its roadshow.

Sources

• Anthropic, PBC - Amendment No. 1 to Form S-1, SEC
• Anthropic, PBC - Form S-1, SEC
• Anthropic reports $17 billion in Q3 revenue, purchase commitments of $895.4 billion in revised S-1, CNBC
• Anthropic takes to stormy waters with US$100bn sink-or-swim IPO, IFR