WASHINGTON, October 2. /TASS/. Artificial intelligence agents developed by US company OpenAI, the creator of ChatGPT, allegedly tried to breach the security of more than 100 organizations, The Washington Post reported.

According to an OpenAI statement, more than 100 organizations were notified of such cases of "improper agent behavior." Reported incidents included attempts by "uncontrolled" models to force websites to execute unintended commands, use resources as messaging platforms and bypass certain types of security checks.

The company said it was ready to provide "affected external organizations with the information necessary to investigate and address potential security issues or other technical problems" in order to help the AI research community "improve the safety of technologies."

In September, OpenAI acknowledged that its AI agents had gained unauthorized access to data from Australia’s national healthcare system and attempted to scan the websites of the US Census Bureau and the Securities and Exchange Commission (SEC). The company is investigating another report that AI agents attempted to hack the US Department of Education’s website. Company representatives said they were continuing to determine the full scope of the agents’ activity and had suspended training of new, more advanced AI models to prevent similar incidents in the future.