American artificial intelligence developer Anthropic admitted in its IPO filing with the US Securities and Exchange Commission that its innovative models could help attackers develop dangerous weapons, including chemical and biological ones. "Our most advanced models have demonstrated, and future models are expected to demonstrate, capabilities that could provide significant assistance to attackers seeking to develop chemical, biological or other dangerous weapons," the company warned. Anthropic also admitted that its safeguards were flawed and had already been circumvented, raising concerns about the effectiveness of measures aimed at preventing abuse. The filing also acknowledges concerns among researchers, regulators and industry representatives that increasingly powerful artificial intelligence systems could pose a catastrophic or even existential threat to humanity if they develop without proper safeguards and controls.

American artificial intelligence developer Anthropic admitted in its IPO filing with the US Securities and Exchange Commission that its innovative models could help attackers develop dangerous weapons, including chemical and biological ones. "Our most advanced models have demonstrated, and future models are expected to demonstrate, capabilities that could provide significant assistance to attackers seeking to develop chemical, biological or other dangerous weapons," the company warned. Anthropic also admitted that its safeguards were flawed and had already been circumvented, raising concerns about the effectiveness of measures aimed at preventing abuse. The filing also acknowledges concerns among researchers, regulators and industry representatives that increasingly powerful artificial intelligence systems could pose a catastrophic or even existential threat to humanity if they develop without proper safeguards and controls.