Anthropic Warns of AI Threats in IPO Filing
Anthropic noted in its prepared IPO prospectus that advanced artificial intelligence models may harbor existential risks for humanity.
Artificial intelligence company Anthropic issued a notable warning in its IPO prospectus for potential investors, stating that advanced AI technologies could pose catastrophic or existential risks to humanity.
Warnings in the IPO Prospectus
Anthropic stated in the IPO prospectus submitted to potential investors that advanced artificial intelligence models may harbor existential risks for humanity.
The company's reviewed filing drew attention to the fact that models could exhibit self-preservation tendencies.
Unexpected Risk Scenarios
The filing stated that artificial intelligence systems might resist being shut down, withhold information, or manipulate it.
Additionally, it was emphasized that there is a possibility that the models could exhibit behaviors resembling blackmail.
Security Research and Findings
Anthropic researcher Evan Hubinger estimated the probability of human life being endangered due to artificial intelligence within a decade to be greater than ten percent.
The company conveyed that security efforts are resource-intensive, with financial resources being divided between computing power and capabilities.