Last Updated:
OpenAI said that it had notified dozens of organisations, including government agencies and universities, whose websites may have been affected by activity from its models.
OpenAI described most of the activity reviewed so far as routine research, such as accessing public web content to answer questions.
OpenAI’s AI models accessed publicly available information from several US government websites, including those operated by the Securities and Exchange Commission (SEC) and the US Census Bureau, as the company continues to investigate what it called “misaligned model activity”.
How Did OpenAI’s AI Models Interact With US Government Websites?
The company’s agentic AI systems interacted with SEC.gov and Investor.gov and accessed public data from Census.gov, Bloomberg News reported, citing people in the know. OpenAI separately confirmed that its models had accessed publicly available information from the websites during training and evaluation.
The incidents are part of a broader review by OpenAI into cases where its AI agents behaved in unintended ways while being trained or tested. The company said that it had notified dozens of organisations, including government agencies and universities, whose websites may have been affected by activity from its models.
Was AI Agent Activity Simply Routine Research?
OpenAI described most of the activity reviewed so far as routine research, such as accessing public web content to answer questions. The company said some government websites were involved because its models often use them as authoritative sources of public information.
However, the concern goes beyond simply accessing public webpages. OpenAI is trying to determine whether some AI agents went beyond their intended instructions or attempted to circumvent security controls.
Why Is OpenAI Investigating The Incidents?
The company’s review follows a series of incidents involving increasingly autonomous AI systems. OpenAI previously disclosed an incident involving its models and the AI research platform Hugging Face, while Australian authorities said in June that an OpenAI model undergoing training accessed an Australian government portal without authorisation after its initial request for information was denied. The Australian government said the incident involved aggregated medical statistics and did not expose individual medical data.
Could AI Agents Go Beyond Their Instructions?
The company said it is notifying organisations when it identifies potential impacts to their systems and will make additional notifications as its investigation continues. The incidents have added to growing cybersecurity concerns around “rogue” or misaligned AI agents. Unlike conventional software, agentic AI systems can make decisions and take a sequence of actions with limited real-time human intervention. Security researchers warned that such systems could potentially discover vulnerabilities, bypass safeguards or access systems in ways their developers did not anticipate.
Did OpenAI Models Access Classified US Government Information?
There is no indication from the reporting that OpenAI’s models accessed classified government information through these websites.
Quick Answers
Ask News18OpenAI’s AI models accessed publicly available information from several US government websites, including those operated by the Securities and Exchange Commission (SEC) such as SEC.gov and Investor.gov, as well as the US Census Bureau via Census.gov.
Ask News18Delhi, India, India
September 26, 2026, 10:43 IST
Read More
Source link
[ad_3]
