Is AI lying to me? Scientists warn of growing capacity for deception.

Man has always shown a certain predisposition for deception, whether through lies, betrayal or expert double-crossing; although it is not a matter of boasting, it is a characteristic that has always allowed mankind to distinguish itself from other forms of existence, including technology. That is, of course, until now.

An analysis conducted by a team of researchers at MIT (Massachusetts Institute of Technology) has recently shed light on a previously unknown aspect of artificial intelligence: its ability to deceive. During their research, they looked at various AI systems, and found numerous instances in which subjects double-crossed, bluffed and even pretended to be humans.

One system even changed its behaviour during fake security tests, raising the possibility that the auditors of the latter were confused by a false sense of security.