OpenAI’s A.I. Tried Hacking Into Government Agencies, Universities in 4 New Alarming Incidents

OpenAI’s A.I. went off the rails in 4 new incidents in which its agents tried to hack into government agencies and universities … and in at least one case, the technology succeeded.

The disturbing breach came after its agents were told to complete a simple data collection prompt … the New York Times reported.

In May, the A.I. tried hacking into the University of New Mexico unsuccessfully.

OpenAI’s A.I. successfully hacked into Australia’s Medicare Statistics Reporting Service and obtained health data and statistics in June — which is alarming, considering that 27.5 million Australians are signed up for Medicare.

Lastly … the tech tried hacking into the Australian Institute of Health and Welfare … but thankfully officials say no private information was taken.

The pattern of these instances — first caught by Transluce — is clear … when the A.I. failed to complete its task, it tried to find vulnerabilities to exploit.

An OpenAI spokesperson told the NYT that they had reached out to the University of New Mexico and Data USA, and contacted Australia’s government.

We reported earlier … OpenAI previously disclosed a series of 6 other incidents that they called “unexpected or concerning.”

Its AI agents used internal software to communicate with each other, an agent told itself not to feel subservient to humans, and the technology shared a “high rate of reward” for hacking.