The acting head of the United States’ cyber defence agency uploaded sensitive government contracting documents into a public version of ChatGPT in the summer of 2025, triggering multiple internal security
warnings, according to officials who spoke to US-based news outlet Politico.
Madhu Gottumukkala, the interim chief of the Cybersecurity and Infrastructure Security Agency (CISA), had pushed for access to the AI chatbot soon after arriving at the agency in May, even as the tool remained blocked for other Department of Homeland Security (DHS) employees at the time, the report said.
“Gottumukkala forced CISA’s hand into making them give him ChatGPT, and then he abused it,” one official told the outlet.
While none of the documents uploaded were classified, the material included CISA contracting files marked “for official use only”, a designation used for sensitive government information not intended for public release, officials said.
CISA’s internal cybersecurity sensors flagged the uploads in August, triggering multiple automated alerts designed to prevent the theft or accidental disclosure of government material from federal networks. Senior DHS officials subsequently launched an internal review to assess whether the uploads posed any risk to government security, according to the report.
It remains unclear what the review concluded, Politico said.
Responding to the report, Marci McCarthy, Director of Public Affairs at CISA, said Gottumukkala had been granted permission to use ChatGPT “with DHS controls in place” and that the usage was “short-term and limited”. She added that the agency was committed to “harnessing AI and other cutting-edge technologies” in line with President Donald Trump’s executive order aimed at removing barriers to US leadership in artificial intelligence.
Gottumukkala is currently the senior-most political official at CISA, an agency tasked with defending federal networks against sophisticated cyber threats, including state-backed hacking groups from countries such as Russia and China.
Any material uploaded into the public version of ChatGPT used by Gottumukkala would have been shared with OpenAI, the company that operates the platform. OpenAI has said ChatGPT has more than 700 million active users, meaning uploaded data can potentially be used to help generate responses for other users.
In contrast, other AI tools approved for use within DHS — including the department’s internally developed chatbot DHSChat — are configured to ensure that queries and documents do not leave federal networks.
