Anthropic (ANT.US) is about to go public, and its prospectus warns of retaliation risk from the Trump administration.

AI company Anthropic PBC (ANT.US) is about to hold a landmark initial public offering (IPO). On the eve of listing, the company warned investors in its prospectus that because it has previously been outspoken on AI safety issues, it could become a target of government retaliation. This highlights that, one year after friction with the Trump administration, the company's future business is facing unusual government-related risks. Prospectus warning: may be "selectively targeted" by the government Anthropic said in its prospectus filed on Monday that the company's positions on AI regulation and remarks by executives may expose it to enormous pressure from the government. The prospectus said: "We have experienced, and may continue to experience, heightened attention, and may continue to face stricter scrutiny from specific government agencies, or the risk of being disproportionately and selectively targeted, even if other similarly situated companies do not encounter the same treatment." These statements highlight the tension between the company and the Trump administration. The tension began earlier this year over a dispute about whether safeguards should be required when the Pentagon uses the company's technology. More recently, Anthropic clashed with officials over whether its frontier model Mythos is safe enough to be publicly released. In addition, CEO Dario Amodei supports coordinated slowing of AI development so that safety teams can keep pace with the technology, a position that has continued to displease U.S. President Donald Trump. Investors face pricing dilemma: regulatory uncertainty hard to measure Investors must now decide how to weigh the regulatory risks facing the company. Anthropic's technology is both close to everyday consumers and can be packaged as everyday products, yet powerful enough to be regarded as a cyber weapon and a potential catalyst for biological warfare, thus requiring stricter scrutiny. For Anthropic, repeated clashes with the government have further intensified the uncertainty hanging over the entire AI industry. Sarah Kreps, director of the Technology Policy Institute at Cornell University, said: "In AI, the boundaries of government power are still being clarified, sometimes through direct confrontations with individual companies. It is difficult for investors to price this uncertainty." In the prospectus, Anthropic disclosed a variety of potential government-related setbacks and mentioned federal legislative proposals, including forced shutdown of models and even nationalization of AI businesses. According to reports, earlier this year, some officials, including Vice President Vance, proposed using the Cold War-era Defense Production Act to force companies to disclose information and cooperate with national security agencies. Anthropic also listed other risks through which government intervention could affect its AI development work, including trade controls that could cut off ties with foreign customers and legal consequences that could arise if its models misbehave. The company said it may not be able to "fully understand or explain" the behavior of the systems it is developing and will continue to push for regulation that could hinder its own business. The company said: "If the regulation or other measures we advocate are adopted, they will apply to us as well, potentially increasing our compliance burden, limiting the flexibility of our product development or deployment, affecting the manner and timing of our development and release of new models, or otherwise slowing our growth." Anthropic also warned that the technology it is developing could ultimately pose a threat to humanity, consistent with the "AI doomsday" concerns of many Silicon Valley AI researchers, which Trump and his allies have dismissed as a "hoax." Anthropic said: "Advanced AI systems, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, could pose catastrophic or existential risks." Anthropic said recent efforts to connect AI models with scientific research could trigger restrictions aimed at preventing the creation of lethal biological weapons. The company added that misuse of its models for cyberattacks could expose the company to legal liability and reputational damage. Behind the disputes: the government is both regulator and customer As the Pentagon dispute highlighted, the government is not only a regulator but also an important customer of the company's products, which further subjects it to the government's will. The company said that during the various periods covered by the filing (dating back to 2023), direct contracts from government agencies accounted for about 1% of its revenue. William Reinsch, former deputy undersecretary for export controls at the U.S. Bureau of Industry and Security, said: "Anthropic is a bit special because in the Trump administration, many actions are taken on a whim, and Anthropic has not received the same treatment as other companies." The Pentagon dispute was triggered by Anthropic's insistence on imposing additional safeguards on its technology and ultimately led U.S. Defense Secretary Hegseth to announce in March of this year that the company posed a supply chain risk and to bar it from military work. Although Anthropic has filed a lawsuit seeking to overturn that determination, a Washington appeals court upheld the Pentagon's plan to remove Claude from its systems and require defense contractors to do the same. Other civilian government agencies are not bound by Hegseth's decision. Anthropic said in the filing: "We cannot determine whether the outcome of these lawsuits will affect our ability to sell products to government agencies and to customers that provide products and services to the U.S. government, and thereby adversely affect our business, financial condition, and operating results." Anthropic has also faced additional scrutiny from the Trump administration over its frontier model Mythos. The company warned in April this year that the model is extremely good at finding computer vulnerabilities and therefore can only be shared with a small number of trusted partners. This breakthrough prompted U.S. Treasury Secretary Bessent to brief Wall Street firms on potential hacker risks and ultimately led Trump to sign an executive order on June 2 establishing a voluntary evaluation system for new frontier models. Because of disagreements with the government over safeguarding the new system, the U.S. Commerce Department briefly imposed export controls on Mythos 5 and a related consumer-facing product, Fable 5, forcing the company to temporarily shut down all access. Two weeks later, after Anthropic resolved a jailbreak concern that could have allowed Fable users to access Mythos capabilities, the export restrictions were lifted. But Anthropic warned in the prospectus that future export controls may not be lifted on a similar timeline and that its future growth will "require" international expansion. Since then, the U.S. government has increased its focus on AI safety. Last month, Trump summoned more than 20 Silicon Valley leaders to the White House and urged AI developers, including Amodei, to sign voluntary agreements, calling on companies to modify board structures, add dedicated safety teams, and conduct independent audits of models. A few days later, Trump administration officials issued legal threats through the Federal Trade Commission, suggesting that companies deemed to have violated the agreements could face consequences. The move sparked concerns about government intervention. U.S. officials, especially Trump administration officials, are increasingly inclined to take more measures to constrain AI companies, which is itself unprecedented. For a company at the center of most of the relevant debates and going public at a time when the United States is trying to regulate an industry still in its infancy, the situation is even more unusual. Matt Pearl, former director of the White House National Security Council and now leading the Strategic Technology Program at the Center for Strategic and International Studies, said: "We don't know what kind of institution Anthropic and other leading AI companies will ultimately become. I have never seen a situation where what a major company will become in the future, and what relationship it will form with the government, is so undetermined."