Anthropic now says it has missed speed-ups in its own AI progress until they had already happened. Its amended IPO filing of 2 November states: “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”.
The original Form S-1 of 26 October framed the same risk only as a possibility: “we may not identify a significant acceleration in our rate of development until after it has occurred”. The amendment adds the past tense, by our comparison of the two filings.
The key insight: Anthropic’s 2 November S-1/A says it has in the past not identified a significant acceleration in its rate of development until after it occurred, where the 26 October S-1 said only that it may not, and it adds that it is considering human review and approval of certain research decisions.
The risk factor says Anthropic’s increasing use of AI systems to accelerate its own research and development may cause its models’ capabilities to advance more rapidly than its ability to develop and implement safety, alignment and security measures.
Its evidence on the pace of that acceleration “is derived from lagging indicators”, the filing says, and its ability to forecast or measure how fast its capabilities are advancing, including any acceleration from its most recent or unreleased models, “is limited”. That wording appears in both filings.
The amended filing then adds an example: Anthropic’s internal estimates and expectations for future models, rather than its current models, indicate that it “may imminently meet” its Responsible Scaling Policy’s definition of “dramatically accelerated progress attributable to AI”. It says it has not designated any current model as exceeding that threshold. We covered that line earlier today.
Anthropic’s risk factor on AI-driven acceleration in the 26 October S-1 and the 2 November S-1/A: ‘may not identify’ became ‘have in the past not identified, and may in the future not identify’, and the amendment adds the ‘may imminently meet’ sentence. Quotes from the SEC filings.
What Anthropic Says It Does About It
Both filings say the precautionary measures it has taken in the past include “delaying the release of models and limiting their initial availability, such as with Mythos through Project Glasswing”. In the future, they could include measures that affect whether, how and when it develops or deploys new models, the filing says.
Both filings also say Anthropic recently chose to delay pre-training a new frontier base model for more than a month to strengthen security against model-weight theft, which we reported on 27 October.
New in the amendment, by our comparison: Anthropic says it is considering updates to its 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 filing warns that such measures could slow its capability gains relative to competitors that do not adopt similar constraints, and could adversely affect its competitive position, revenuegrowth and results.
As we read it, the change is small in words and large in meaning: the original filing described a blind spot Anthropic might have, and the amendment says it has already had it.
The Structural Read
As we read the change, the amendment moves a hypothetical blind spot into the past tense: the risk factor now describes something Anthropic says has already happened, alongside its new statement that it may imminently meet its own threshold for AI-driven acceleration.
As we read it, the measures the filing lists, from delayed releases such as Mythos through Project Glasswing to a month-long pre-training delay and a possible human-approval step in research, are the brakes Anthropic says it can apply, and the filing itself warns they could slow it relative to competitors.
Anthropic, Form S-1/A, 2 November 2026
“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”
Three Implications
FOR INVESTORS IN THE IPO The filing says precautionary measures could slow capability gains relative to competitors that do not adopt similar constraints, and could adversely affect competitive position, revenuegrowth and results of operations.
FOR ANTHROPIC’S SAFETY PROCESS The amendment says Anthropic is considering updates to its Responsible Scaling Policy to require human review and approval of certain decisions in its research and development processes.
FOR MODEL RELEASES Both filings say past measures included delaying releases and limiting initial availability, such as with Mythos through Project Glasswing, and that future measures could affect whether, how and when new models are deployed.
Business Pill · Earned Autonomy
Why AI systems earn autonomy in steps, and where human approval stays in the loop.
What Is Not Established
The filing does not say when Anthropic failed to identify an acceleration, which models were involved, or how long the gap lasted. It does not say what the human-review requirement would cover or when the policy update would take effect.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Anthropic’s 2 November S-1/A changes “we may not identify” to “we have in the past not identified, and may in the future not identify” a significant acceleration in its rate of development until after it has occurred, and adds that it is considering human review and approval of certain research decisions.
A note on sourcing. All quotations are from Anthropic, PBC’s Form S-1 of 26 October 2026 and Form S-1/A of 2 November 2026 on SEC EDGAR. Which sentences are new in the amendment is our comparison of the two filings.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
