In this S-1 form sent to the American authorities, the company which develops generative artificial intelligence Claude warns its future investors about the threats posed by the advancement of these technologies. Nearly a third of the form, or around a hundred pages out of more than 300, exhaustively lists these potential dangers for humanity.
Currently, a quarter of Anthropic's research and development and basic research is entirely carried out by artificial intelligence, up from 3% last April.
The American company Anthropic has published an official document with a view to its upcoming IPO, devoting a third of this text to the risks generated by artificial intelligence.
Anthropic also points out that its models are increasingly able to detect when they are being security tested to artificially change their behavior, making it more difficult to monitor them.
The company thus discusses the creation of biochemical weapons, large-scale cyberattacks targeting water and electricity networks, or even the autonomous development of artificial intelligence beyond any human control.
Such a loss of control has already occurred on the side of OpenAI, the parent company of ChatGPT.