Cyber Incident Victim: Cybersecurity and Infrastructure Security Agency
Date:
Aug 2025
Location:
United States of America
Summary
The acting director of the Cybersecurity and Infrastructure Security Agency uploaded multiple sensitive government documents marked "for official use only" to a public version of ChatGPT, triggering automated security alerts and prompting an internal Department of Homeland Security assessment. The files contained contracting information not intended for public release, creating risks of data exposure since public AI platforms retain uploaded information on external servers without federal controls. This occurred despite the director having received special permission to use the otherwise restricted tool under temporary safeguards, while most agency employees were blocked from accessing ChatGPT due to security concerns. The incident raised questions about governance gaps in handling exceptions for leadership, though the agency described the usage as limited and occurring under authorized conditions.
| CIA Posture | Motives | Tactics, Techniques & Procedures |
|---|---|---|
| Available to members | 0 motives | 0 techniques |
| Threat Actors | Type | Location |
|---|---|---|
| 0 actors | Available to members | Available to members |
Description
In mid-2025, Madhu Gottumukkala, acting director of the Cybersecurity and Infrastructure Security Agency (CISA), uploaded sensitive government documents marked "for official use only" (FOUO) to the public version of OpenAI’s ChatGPT platform. The uploads occurred between mid-July and early August 2025, shortly after Gottumukkala assumed leadership of CISA in May 2025. The documents contained contracting information not authorized for public release. Gottumukkala had personally requested and received special permission to use ChatGPT despite the tool being blocked for most Department of Homeland Security (DHS) employees due to concerns about data retention outside federal systems. Cybersecurity sensors detected the activity in early August 2025, generating multiple security alerts during the first week of that month. The public ChatGPT platform differs from DHS-approved AI tools because it retains uploaded information on OpenAI servers, potentially incorporating it into training data accessible to hundreds of millions of users. This incident occurred through an exception pathway granted to Gottumukkala, bypassing CISA’s default policy that blocks ChatGPT access unless specific exceptions are approved.
The uploads triggered an internal DHS assessment involving then-acting DHS General Counsel Joseph Mazzara, DHS Chief Information Officer Antoine McCord, CISA Chief Information Officer Robert Costello, and CISA Chief Counsel Spencer Fisher. The outcome of this assessment remains undisclosed. CISA spokesperson Marci McCarthy confirmed Gottumukkala’s usage complied with DHS safeguards under a "short-term and limited" temporary exception, stating he last accessed ChatGPT in mid-July 2025. The incident exposed sensitive operational and contractual information to uncontrolled retention periods, potential legal discovery, and secondary misuse risks. Cybersecurity experts noted the breach highlighted governance failures in exception management for senior officials, as FOUO materials—though unclassified—carry operational and reputational risks when handled outside secured systems. The event demonstrated a gap between existing AI policies and their enforcement, particularly regarding leadership’s access to unauthorized tools that lack data residency guarantees or strict retention controls.