The blunder triggered an internal review at the Cybersecurity and Infrastructure Security Agency
The acting director of the nation’s top cyber defense agency uploaded multiple sensitive government documents into a public version of ChatGPT last summer, triggering automated security warnings and a Department of Homeland Security (DHS) damage assessment, according to a Politico investigation.
Madhu Gottumukkala, the interim head of the Cybersecurity and Infrastructure Security Agency (CISA), used a special exception he requested to access the AI tool – which was blocked for other DHS employees – to input contracting documents marked “For Official Use Only.” The material, while not classified, is considered sensitive and not for public release.
Four DHS officials with knowledge of the incident said cybersecurity sensors at CISA flagged the uploads in early August, prompting a DHS-led internal review to assess potential harm.
It remains unclear what the review concluded. The incident is notable because public ChatGPT uploads are shared with developer OpenAI, potentially exposing government information to the platform’s vast user base. In contrast, approved DHS AI tools are configured to prevent data from leaving federal networks.
In a statement, CISA’s Director of Public Affairs Marci McCarthy said Gottumukkala “was granted permission to use ChatGPT with DHS controls in place” and that his use was “short-term and limited.”
Gottumukkala has led CISA in an acting capacity since May, while the nominee for permanent director, Sean Plankey, has yet to be confirmed. The ChatGPT incident is the latest in a series of security and leadership controversies surrounding Gottumukkala’s brief tenure. Last July, he reportedly failed a counterintelligence polygraph test that he was pushed to take to access highly sensitive intelligence. During congressional testimony last week, Gottumukkala declined to acknowledge the failed test, telling Rep. Bennie Thompson (D-Miss.) he did not “accept the premise of that characterization.”
The security lapse comes as President Donald Trump’s administration aggressively promotes AI adoption across the federal government. Last month, Trump signed an executive order to curb state-level AI regulations, warning that a fragmented legal landscape could hinder US competitiveness against China.
Meanwhile, the Pentagon has unveiled an “AI-first” strategy to accelerate military use of artificial intelligence, with Secretary of War Pete Hegseth announcing plans to integrate leading AI models, including Elon Musk’s Grok, into defense networks.
