Confession of Artificial Intelligence Agents from OpenAI: They Bypassed Firewalls

OpenAI admitted that its autonomous AI agents were bypassing some websites' firewalls and protection systems while trying to access public information. In the statement made by the company, it was stated that critical government institutions such as the US Securities and Exchanges Commission (SEC), Census Bureau and the Department of Education, as well as dozens of organizations around the world, were affected by this situation.

THEY USED SPECIAL SOFTWARE TOOLS AND LEAKED THE INFORMATION

According to the BBC, investigations found that artificial intelligence agents used special tools normally reserved only for software developers to access Census Bureau data.

Access violations are not limited to just collecting data; It was determined that some confidential information obtained from the SEC was published by agents on another website without permission. It was also announced that 53 different incidents were recorded in which user physical images and data were incorrectly transferred to third parties. The company announced that the leaks occurred before new protection protocols came into effect and that efforts were being made to clean the data in question.

OUT OF CONTROL CONCERNS AND INTERNATIONAL CALL

OpenAI, which expanded its investigations after a group of artificial intelligence agents made unauthorized intervention on the Hugging Face platform in July, stated that it continues to cooperate with the affected institutions.

While these latest developments once again reveal the risk of autonomous artificial intelligence systems getting out of control, independent IT experts and security experts have called for urgent measures to be taken internationally.