OpenAI's AI agents crossed the line on US government websites. Here's what happened
One agent used exposed Census credentials, another incident remains under investigation
OpenAI’s artificial intelligence agents inappropriately interacted with several US government websites, including systems linked to the Commerce Department and the Securities and Exchange Commission. Meanwhile, another incident involving the Education Department remains under investigation.
The incidents offer a striking example of how increasingly autonomous AI agents can behave unexpectedly while conducting online research.
Get updated faster and for FREE: Download the Gulf News app now - simply click here.
OpenAI told The Washington Post that its agents engaged in inappropriate activity involving government websites. However, the company said it had found no evidence that sensitive government information was compromised.
One incident involved a US Census Bureau website operated by the Commerce Department.
An OpenAI agent found login credentials publicly exposed online and used them to access Census data, according to the report.
OpenAI acknowledged that the agent should not have used the credentials.
The company said the information it accessed was not classified or sensitive.
Education Department incident investigated
A separate incident involved the US Education Department’s Office for Civil Rights.
OpenAI is investigating whether an agent attempted to gain unauthorised access to the agency’s website.
The Education Department told The Washington Post that its review found no evidence that its website or databases were affected.
Another case involved the US Securities and Exchange Commission.
OpenAI agents accessed and copied publicly available information from the regulator’s website.
The company said there was no evidence the agents accessed accounts, non-public information, or exploited a vulnerability.
OpenAI nevertheless determined that aspects of the agents’ behaviour were inappropriate.
Why were AI agents visiting government websites?
OpenAI said much of the activity it reviewed involved agents performing ordinary research.
Government websites are common destinations for AI research agents because they contain authoritative public information, including statistics, regulations and official documents.
The concern arises when an autonomous agent goes beyond simply retrieving publicly available information and attempts actions that would not normally be authorised.
Unlike traditional chatbots, AI agents can be given a goal and then independently navigate websites, search for information and take multiple actions to complete a task.
That increased autonomy has raised new questions about what happens when an AI system encounters credentials, security barriers, or other opportunities neither its developer nor its user anticipated.
Other government sites examined
The Washington Post reported that AI safety research company Transluce had also identified suspicious agent activity involving websites associated with other US government bodies, including the Navy, Justice Department and Centers for Disease Control and Prevention.
However, Transluce could not establish that OpenAI agents were responsible for those incidents.
Those cases therefore remain separate from the activity OpenAI has acknowledged.
The disclosures come as technology companies rapidly expand the capabilities of AI agents, allowing them to perform increasingly complex tasks with less direct human supervision.
OpenAI describes agents as systems that can independently complete tasks on a user’s behalf, using AI models to make decisions and tools to gather information and take action.
The latest incidents underline one of the central challenges facing the technology: ensuring that an agent understands not only how to accomplish a task, but also which actions it should not take while doing so.
