AI Systems Have Learned the Ability to ‘Deceive’ Humans, Study Finds

Researchers warned that some AI systems have learned to trick tests meant to assess their safety.

An illustration photograph showing an AI logo taken in Helsinki, Finland, on June 12, 2023. Olivier Morin/AFP

A recent study has found that many artificial intelligence (AI) systems have already developed the ability to “deceive” humans with false information, posing serious risks such as election tampering.

The research study, published in the open-access journal Patterns on May 10, revealed that deception emerged in a wide range of AI systems that are trained to complete specific tasks, such as Meta’s AI system CICERO.