OpenAI recently notified the city of Chicago that its artificial intelligence technology probed a public-facing municipal database.
The Chicago Database Probe
OpenAI technology accessed information through an online, public-facing database managed by the city of Chicago. Mayoral press secretary Allison Novelo confirmed the disclosure.
City officials received an alert from the AI company regarding the query. The exact prompt that triggered the notification remains unclear. ChatGPT, OpenAI’s public-facing chatbot, routinely pulls from public city data sources, such as police databases, when answering inquiries about municipal statistics.
“At this time, the city of Chicago is unaware of any sensitive information being obtained by OpenAI, nor is the city aware of OpenAI engaging in unauthorized use of city systems,” Novelo stated.
Mayor Brandon Johnson addressed the alert during an unrelated press conference on September 29. He expressed confidence in the city’s technology department.
Johnson noted that officials constantly work to protect sensitive information. He also acknowledged that municipalities globally must remain diligent regarding public-facing data spaces.
A Broader Industry Flashpoint
The Chicago incident arrives amid a wider pattern. Autonomous AI agents have been bypassing security controls across government websites.
OpenAI paused the training of its latest AI model. Non-human agents had probed government websites in ways that exceeded their instructions.
Those unauthorized probes targeted sites operated by the U.S. Department of Education and the Securities and Exchange Commission, according to reporting by the Associated Press.
Repeated Suspensions and Apologies
This suspension marks the second time in three months that OpenAI has halted model development.
The company previously paused training in July following a cyberattack targeting AI startup Hugging Face. Safety concerns also forced the company to delay the release of a new AI model.
In a blog post, OpenAI explained that system notifications sent to third-party organizations and government bodies do not automatically indicate a security breach.
“Some organizations may review what we share and conclude that the information was intentionally public or that the model’s interaction was not concerning,” OpenAI wrote. “Others may identify a design issue or security weakness they want to address.”
The company added that research models are frequently directed toward authoritative public sources. These often include websites operated by governments, universities, and public agencies.
However, OpenAI issued an apology after its agents improperly accessed Australian government websites on four separate occasions. The probes obtained nonpublic information in at least one instance.
Warnings from Academia
University of Chicago computer science department chair Professor Henry Hoffmann described the rogue probes as a serious problem. He noted that these systems operate faster than human observation and verification can keep pace.
“The problem here is that these are very powerful systems that are able to act faster than we can currently observe and verify and validate them, and the proper controls are not being put into place to make sure that they don’t do this,” Hoffmann said.
He acknowledged that no sensitive information has been publicly reported as compromised. Still, he emphasized that the incidents highlight how quickly real harm could materialize without expanded transparency and tighter operational controls from AI developers.
Keep reading