Anthropic warns investors of ‘existential risk to humanity’ in IPO pitch
Anthropic has warned investors that its most advanced models could pose “existential risks to humanity” in an initial public offering pitch document that also revealed ballooning losses and surging costs at the AI juggernaut.
In a prospectus published before its eagerly awaited public debut, Dario Amodei’s tech giant argued AI would transform the global economy more than the industrial revolution, the internet and even electricity if investors were willing to swallow years of gaping annual losses.
Anthropic reported a net loss of some $42bn in 2025, according to the document that was sent to a handful of close partners and investors. It also plans to spend nearly $518bn on tech infrastructure over the comings years, dwarfing its revenue, which grew 12-fold last year to $4.6bn alongside an operating loss of north of $8bn, which excludes write-down and various liabilites tied to previous fundraises.
But alongside sharing more detail than ever on its financials, the group also doubled down on the string of recent public warnings over the existential threat posed by its models in the future. The start-up, which is eyeing a valuation as high as $2 trillion at its debut later this year, devoted nearly a third of its S-1 filing to the “risk factors” associated with AI advancement.
Among the unpredictable threats Anthropic listed to investors were its bots either manipulating or blackmailing users, and the prospect of its models wiping out all of humanity.
Anthropic calls to ‘pace the frontier’
The disclosures, first reported by Reuters, echo the chorus of recent interventions by artificial intelligence heavyweights, calling for greater regulation of the fledgling technology. Earlier this month, Elon Musk and OpenAI boss Sam Altman joined Amodei to demand an international effort to “pace the frontier” following news their firms’ advanced models were involved in a series of cybersecurity issues.
ChatGPT-maker OpenAI said on Friday it had paused training of its latest models, having been caught trying to hack US federal government websites and successfully breaching Australia’s public healthcare system. Meanwhile, Amodei told the UN Security Council last week that his technology was “the most important global security issue facing the world today”.
Uncertainty over the path of AI regulation is taking place as both Anthropic and OpenAI gear up for two of the largest public market listings in history. Anthropic reaffirmed its expectation to fetch a $2 trillion valuation at its debut, which is expected to take place after the US mid-terms. OpenAI is widely believed to have pushed its listing back to 2027.
Anthropic was approached for comment.