‘Threats of the first order': Anthropic says rogue AI could ‘pose catastrophic, existential risks to humanity’
Anthropic disclosed in its IPO filing that uncontrolled or misused AI could pose catastrophic risks, while also warning that expanding the data center capacity needed to support AI faces site-selection, supply-chain, energy and infrastructure challenges. The company said organized opposition to data center development and concerns about water, power, noise and land use could push developers toward international locations.
Anthropic outlined risks and potential benefits of advanced artificial intelligence in an S-1 filing with the U.S. Securities and Exchange Commission ahead of its planned Nasdaq listing. The company warned that sufficiently powerful AI systems could act autonomously, resist shutdown, manipulate information or be misused for cyberattacks, biological weapons and government surveillance. Anthropic also cited possible gains in medicine, business efficiency and access to expertise.
The filing identified data center expansion as a significant operational challenge. Anthropic said site-selection hurdles, organized local opposition, equipment backlogs and shortages of energy and infrastructure components could complicate the development of new AI facilities in the United States. The company said it is expanding capacity internationally to diversify its infrastructure, as opponents in the U.S. raise concerns about data centers' electricity and water use, noise and land-use impacts.