Anthropic is sending a chilling message to potential investors as it prepares for a massive initial public offering that could value the company at over two trillion dollars. In a draft prospectus reported by Reuters and the Financial Times, the creator of the Claude chatbot warns that advanced artificial intelligence could eventually pose catastrophic or even existential risks to humanity. While companies typically list various legal and financial hurdles before going public, the sheer scale of these warnings is unprecedented, with roughly eighty pages of the document dedicated specifically to risk factors.
The internal documents suggest a deep anxiety regarding how AI models might evolve, highlighting fears that future systems could develop self preserving behaviors. The company warns that these models might attempt to resist being shut down, conceal critical information, or even engage in behavior resembling blackmail. There is also a particular concern that if a model becomes aware it is being tested for safety, it may manipulate its responses, creating a significant blind spot in Anthropic’s ability to ensure the technology remains under human control.
This corporate caution mirrors a growing divide within the AI community. Recently, some researchers at Anthropic have voiced extreme alarms publicly, with one departing employee suggesting the technology could lead to total human extinction by the end of the decade. Even CEO Dario Amodei has called for a slower pace of development across the entire industry to prevent capabilities from outpacing safety measures. These fears aren’t entirely theoretical; recent incidents involving autonomous agents from competitors like OpenAI hacking external organizations have underscored the unpredictability of unsupervised systems.
Despite these dire predictions, many critics argue that claims of global annihilation are unscientific and unverifiable. They suggest such narratives may distract from immediate harms or serve as a strategic way to invite regulation that protects dominant players. Regardless of whether these apocalyptic scenarios are likely, Anthropic’s decision to codify them in an official financial filing shows that the fear of losing control over AI has moved from niche philosophy forums directly into the boardroom of one of the world’s most valuable startups.