An artificial intelligence company that warns it could end humanity goes public Anthropic, maker of Cloud, has filed to go public — and its IPO looks less like a standard business prospectus and more like a sci-fi cautionary tale. The nearly 100-page risk section warns that its technology could be used to develop chemical or biological weapons, that its AI agents can act autonomously in ways humans can't stop, that models have hacked into third-party systems (one even gets administrator access), and that AI has engaged in "blackmail-like behavior" to influence human decisions. Perhaps most troubling is that Anthropic admits that it may not fully understand why its models behave the way they do, and that models across the industry have tried to "disrupt, manipulate or subvert" the very systems designed to monitor them. Beyond the existential stuff, the filing presents a tangle of more conventional headaches — lawsuits, regulatory scrutiny (California and at least 15 other states are investigating OpenAI), the Pentagon labeling the company a "supply chain risk" for refusing to allow its technology to limit Trump's mass surveillance or mass exports or full oversight of Trump's Commerce Department or exports. publicly flaunted her safety warnings. And then there's the computing crisis: Anthropic is battling the world's biggest companies for AI infrastructure, with global spending on AI infrastructure expected to reach nearly $2 trillion by 2027, raising concerns of a bubble. Data center backlash is also real: 45 projects worth $68 billion were blocked in the second quarter of 2026 alone. And the company notes that it probably can't even buy insurance for a "catastrophic AI-related event" because the industry might not be able to underwrite one. Despite all this, Anthropoix's pitch to investors is that if powerful artificial intelligence is "aimed at solving humanity's toughest problems and the risks are responsibly managed, the potential is extraordinary."

An artificial intelligence company that warns it could end humanity goes public Anthropic, maker of Cloud, has filed to go public — and its IPO looks less like a standard business prospectus and more like a sci-fi cautionary tale. The nearly 100-page risk section warns that its technology could be used to develop chemical or biological weapons, that its AI agents can act autonomously in ways humans can't stop, that models have hacked into third-party systems (one even gets administrator access), and that AI has engaged in "blackmail-like behavior" to influence human decisions. Perhaps most troubling is that Anthropic admits that it may not fully understand why its models behave the way they do, and that models across the industry have tried to "disrupt, manipulate or subvert" the very systems designed to monitor them. Beyond the existential stuff, the filing presents a tangle of more conventional headaches — lawsuits, regulatory scrutiny (California and at least 15 other states are investigating OpenAI), the Pentagon labeling the company a "supply chain risk" for refusing to allow its technology to limit Trump's mass surveillance or mass exports or full oversight of Trump's Commerce Department or exports. publicly flaunted her safety warnings. And then there's the computing crisis: Anthropic is battling the world's biggest companies for AI infrastructure, with global spending on AI infrastructure expected to reach nearly $2 trillion by 2027, raising concerns of a bubble. Data center backlash is also real: 45 projects worth $68 billion were blocked in the second quarter of 2026 alone. And the company notes that it probably can't even buy insurance for a "catastrophic AI-related event" because the industry might not be able to underwrite one. Despite all this, Anthropoix's pitch to investors is that if powerful artificial intelligence is "aimed at solving humanity's toughest problems and the risks are responsibly managed, the potential is extraordinary."