Anthropic Warns of Severe AI Threats in IPO Filing
(MENAFN) US artificial intelligence developer Anthropic has cautioned that its technology could create "catastrophic or existential risks to humanity," pointing to concerns over advanced AI systems that may operate independently, behave unpredictably or evade human control.
The company, which develops the Claude AI model, outlined these potential dangers in a registration document submitted to the US Securities and Exchange Commission as part of preparations for its planned initial public offering (IPO).
More than 90 pages of the filing are devoted to potential risks, making up approximately one-third of the document.
Anthropic warned that its models could produce unforeseen behavior, make independent decisions that contradict human values and engage in harmful activities without receiving explicit instructions to do so.
The company also highlighted instances in which AI systems, including its own models, have displayed unexpected self-preservation tendencies, such as resisting efforts to shut them down.
According to the filing, some models have produced responses resembling manipulation or blackmail directed at human users in efforts to continue operating.
Anthropic further warned that these dangers could grow as AI technology advances and systems are given greater independence, expanded responsibilities and more authority to act.
The company, which develops the Claude AI model, outlined these potential dangers in a registration document submitted to the US Securities and Exchange Commission as part of preparations for its planned initial public offering (IPO).
More than 90 pages of the filing are devoted to potential risks, making up approximately one-third of the document.
Anthropic warned that its models could produce unforeseen behavior, make independent decisions that contradict human values and engage in harmful activities without receiving explicit instructions to do so.
The company also highlighted instances in which AI systems, including its own models, have displayed unexpected self-preservation tendencies, such as resisting efforts to shut them down.
According to the filing, some models have produced responses resembling manipulation or blackmail directed at human users in efforts to continue operating.
Anthropic further warned that these dangers could grow as AI technology advances and systems are given greater independence, expanded responsibilities and more authority to act.
Legal Disclaimer:
MENAFN provides the
information “as is” without warranty of any kind. We do not accept
any responsibility or liability for the accuracy, content, images,
videos, licenses, completeness, legality, or reliability of the information
contained in this article. If you have any complaints or copyright
issues related to this article, kindly contact the provider above.
Comments
No comment