The acting director of CISA uploaded sensitive government contracting documents marked 'for official use only' into the public version of ChatGPT, triggering multiple automated security warnings and an internal DHS review.
In summer 2025, Madhu Gottumukkala, the interim head of the Cybersecurity and Infrastructure Security Agency (CISA), uploaded sensitive contracting documents into the public version of ChatGPT. Gottumukkala had requested and received special permission to use ChatGPT soon after arriving at the agency in May, while the app was blocked for other DHS employees. The documents were marked 'for official use only,' a government designation for sensitive information not intended for public release, though none were classified. Cybersecurity sensors at CISA flagged the uploads in August 2025, with multiple warnings occurring in the first week of August alone. Any material uploaded to the public ChatGPT is shared with OpenAI and can be used to help answer prompts from other users of the app, which has over 700 million active users. Senior DHS officials subsequently led an internal review to assess potential harm to government security, involving the acting general counsel Joseph Mazzara and chief information officer Antoine McCord. Gottumukkala had meetings with CISA's chief information officer Robert Costello and chief counsel Spencer Fisher about the incident and proper handling of sensitive material. The conclusion of the internal review is not specified in the report.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that memorize and leak sensitive personal data or infer private information about individuals without their consent. Unexpected or unauthorized sharing of data and information can compromise user expectation of privacy, assist identity theft, or cause loss of confidential intellectual property.
Human
Due to a decision or action made by humans
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
No population impact data reported.