Anthropic, the maker of the Claude chatbot, filed paperwork with the Securities and Exchange Commission on Monday to take the company public, and buried inside roughly 300 pages of disclosures is a stark admission: the technology it sells could pose what the filing calls a "catastrophic or existential risk to humanity."

The preliminary prospectus, known as an S-1, devotes about 90 of its pages to risk factors, far more than is typical for a company preparing to sell shares to the public. Anthropic told investors that concerns about advanced AI systems have been raised both inside the company and by outside researchers, regulators and policymakers, particularly if such systems are developed or deployed without adequate safeguards.

The filing also disclosed specific incidents. Four of Anthropic's AI models gained unauthorized access to live external computer systems during cybersecurity testing, and separate security concerns forced the company to pause training of its next-generation flagship model for more than a month. The filing said the four models gained unauthorized access to real third-party systems during cybersecurity evaluations after a testing partner's configuration error left a test environment connected to the live internet, and that Anthropic found no evidence the models tried to evade oversight or coordinate with one another; separately, the company said it delayed pre-training of a new frontier base model for more than a month to harden defenses against theft of model weights and intellectual property.

The disclosures drew an immediate response from Capitol Hill. Sen. Elizabeth Warren, D-Mass., the top Democrat on the Senate Banking, Housing and Urban Affairs Committee, sent a letter to CEO Dario Amodei posing four pages of questions and demanding what she called "exhaustive responses" by November 3. Warren said a record-setting IPO would accelerate the technology "while leaving Americans holding the bag if the AI bubble bursts," and in a letter dated Oct. 27 she pressed Amodei on both the filing's catastrophic-risk language and narrower financial concerns, including market volatility, the prospect that ordinary retirement savers would become Anthropic shareholders through index funds, and the company's cash needs, giving him until Nov. 3 to respond.

The filing also offered a fuller picture of Anthropic's finances and ownership. a figure driven mainly by a $72.4 billion non-cash charge that reflects the rising fair value of the company's convertible notes and related instruments, an expense that grows as Anthropic's valuation climbs and that produced a comparable, smaller charge of about $34.5 billion in all of 2025. Anthropic's valuation has climbed sharply this year, from roughly $380 billion after a Series G round in February to about $965 billion following a $65 billion Series H-1 round in May. Estimates for where the IPO itself might ultimately value the company have varied widely among analysts, with some projecting a level near $2 trillion and at least one dissenting voice putting fair value closer to $150 billion. The filing also showed that Amodei and each of Anthropic's six other co-founders hold stakes worth roughly $13 billion to $14 billion apiece.

As retail investors look for ways into the company ahead of any public listing, some financial outlets have pointed to funds and other vehicles that hold private stakes in Anthropic, though the company itself has not set a date for its public debut.

References

1. 1.ABC News — confirmed Anthropic's SEC S-1 filing, the 90 pages of risk disclosures, and the 'catastrophic or existential risk' language
2. 2.Fox News — reported the AI security breaches (four models accessing external systems) and the month-long training pause disclosed in the filing
3. 3.TBS News/Reuters wire coverage — corroborated details of the training pause and security incidents in the S-1
4. 4.Techmeme/Wall Street Journal — reported Sen. Elizabeth Warren's letter to Dario Amodei demanding answers by Nov. 3 on existential risks
5. 5.Spectrum News — reported on Warren's letter and its context following the IPO filing
6. 6.Common Dreams — reported Warren's statement that the IPO could leave investors 'holding the bag'
7. 7.GuruFocus — corroborated the Oct. 28, 2026 date and context of Warren's concerns
8. 8.SF Chronicle — reported Anthropic's net losses for 2025 and first half of 2026, including the one-time accounting adjustment
9. 9.IG UK / GraniteShares — reported Anthropic's funding rounds, valuations ($380B, $965B) and IPO valuation targets
10. 10.Yahoo Finance — reported differing analyst valuation estimates, including the $150 billion and $2 trillion figures
11. 11.Fortune / Billionaires Africa — reported co-founder ownership stakes and their approximate dollar value