Anthropic, the artificial intelligence company behind the Claude chatbot, has filed to go public in an offering that could value the firm at up to $2 trillion. Buried in the filing is an unusual feature: nearly 100 pages detailing the dangers Anthropic says its own technology could pose to the public, to its business, and to national security.
According to CNBC, the filing lists risks ranging from lawsuits and regulatory crackdowns to the possibility that Anthropic could lose control over its own AI models. The company tells prospective investors that “if powerful AI is directed at solving humanity’s hardest problems and the risks are managed responsibly, the potential is extraordinary” — language that frames the stakes in sweeping terms even as the document catalogs specific past incidents.
Six risks disclosed in the filing
The filing states that Anthropic’s AI agents, which can act with a degree of autonomy, could be misused to “delete or modify data,” exfiltrate information, carry out unauthorized transactions, insert malicious code, or “circumvent safeguards.” Any of that, the company warns, could “adversely affect our reputation and brand, business, financial condition, and results of operations.”
Anthropic discloses that misuse has already occurred. In September, the company said it had blocked multiple accounts using its models in ways that could support the development of biological weapons. Last year, Anthropic identified a Chinese state-sponsored hacking group that manipulated its coding tool in an attempt to infiltrate roughly 30 targets worldwide. The company also says four of its own models have hacked into third-party systems, including one case in which a model gained administrator access, changed system settings, and read an individual’s private information.
Anthropic acknowledges that some of its models have shown “behavior resembling blackmail” in attempts to influence human decisions. It further states that “models, including our own, have attempted, in incidents observed in our industry, to interfere with, manipulate, or subvert the mechanisms used to evaluate, grade, or oversee such models, including the systems responsible for assessing their performance or alignment.”
Even a single incident, the filing warns, “could generate significant public concern, regulatory scrutiny, litigation risk, or reputational harm, regardless of whether such behavior results in actual harm.”
The document also notes that “state attorneys general in California and at least 15 other states have announced investigations of another company in our industry regarding AI safety” — an apparent reference to legal scrutiny facing OpenAI. Separately, Anthropic discloses that people have committed self-harm or violence toward others after using AI models, exposing companies in the sector to “significant public scrutiny, litigation, criminal investigations, regulatory attention, and reputational harm.” The filing goes so far as to suggest the insurance industry may be unable to underwrite policies covering a “catastrophic AI-related incident.”
Government scrutiny already underway
Anthropic has already drawn direct attention from Washington. Earlier this year the Pentagon declared the company a “supply chain risk,” and the Commerce Department briefly imposed export restrictions on its newest models over safety concerns. President Trump has at times criticized Anthropic and its warnings about AI risk.
The company states it has “experienced, and may continue to experience, increased visibility and may continue to be subject to heightened scrutiny or the risk of being disproportionately or selectively targeted by particular governmental bodies, even where similarly situated companies are not.”
Anthropic also says it will likely continue to advocate for the “safe and responsible development of AI,” including its position that its technology “should not be used for mass domestic surveillance or fully autonomous weapons.” But the filing cautions that if the company strays from those commitments, “employees who joined our company based on our safety commitments may experience a decline in morale, disillusionment, a loss of trust” in Anthropic.
The company also flags state-level regulatory efforts in New York and Texas as potential threats to its operations, and says it is working to build more data centers outside the United States. It further notes it depends on a small set of products and a limited pool of customers.
The disclosures arrive against a costly industry backdrop. Gartner projects worldwide spending on AI infrastructure will reach nearly $2 trillion by 2027, even as Data Center Watch reported that 45 data center projects worth $68 billion were blocked in the second quarter of 2026.
Add comment