⚠️ DISCLAIMER: Ministry of Cyber Affairs is an independent news, guides and directory service on cybercrime and cybersecurity, for victims and investigators. It is NOT an official government or ministry portal.
Website Scam CheckerNew
Back to News

OpenAI Disclosed ChatGPT User Content, Including Prompts and Files, in 88 Government Requests

12 November 2026
Source:OpenAI
Share
OpenAI Disclosed ChatGPT User Content, Including Prompts and Files, in 88 Government Requests

OpenAI disclosed user content, including text and files entered into or generated by its services, in response to 88 government requests in January–June 2026, covering 130 accounts. The company received 146 content requests, typically search warrants, and produced data in about 60 percent of them, plus 620 non-content requests for subscriber records. Ordinary requests more than doubled from late 2025. No country breakdown is given. U.S. national security demands totaled 250–499.

OpenAI disclosed user content, including text and files that people entered into or received from its services, such as ChatGPT prompts and outputs, in response to 88 government requests in the first half of 2026.

Those disclosures covered 130 user accounts, according to the company's transparency report for January to June 2026.

The company received 146 content requests in the period, described as legal process such as search warrants seeking "text or files input or output in connection with the use of OpenAI services."

It produced data in 88 of those cases, about 60 percent, and did not disclose data in the other 58.

The report refers to OpenAI services generally and does not name ChatGPT, but the definition covers user inputs and model outputs on those services.

Most Requests Sought Account Records, Not Conversations

Content requests were the minority.

OpenAI received 620 non-content requests, legal process such as subpoenas for an account holder's name, email address, payment information and transaction history.

It disclosed data in 441 of those requests, covering 679 accounts.

It also received 28 emergency requests, in which it may release user data if it believes disclosure is necessary to prevent death or serious physical injury.

Data was disclosed in 10 of those cases, affecting 13 accounts.

The report does not say whether emergency disclosures included conversation content.

Excluding U.S. national security process, OpenAI received 794 government and law-enforcement requests and disclosed data in 539 of them, or about 68 percent, covering 822 accounts.

Non-content requests accounted for 78.1 percent of that total, content requests 18.4 percent and emergency requests 3.5 percent.

The report does not break requests down by country, so it is not possible to identify which governments received prompt or other content data.

It also does not separately tally subpoenas, court orders and search warrants.

U.S. National Security Demands Reported Only as a Range

Requests under the U.S. Foreign Intelligence Surveillance Act (FISA) and National Security Letters are excluded from the main table and reported only in ranges required by U.S. law.

From January to June 2026, OpenAI received 250 to 499 such demands, affecting 250 to 499 accounts.

FISA orders can seek both non-content and content information.

National Security Letters are limited to information about a user's identity.

OpenAI did not say how many of these demands led to disclosure, or whether any produced prompts or other content.

Requests More Than Doubled From Late 2025

The ordinary request totals are a sharp increase from the previous six months.

In July to December 2025, OpenAI received 224 non-content requests, 75 content requests and 10 emergency requests, 309 in all, and disclosed content in 62 cases covering 84 accounts.

U.S. national security process in that period was reported in the lower 0-to-249 range.

In the first half of 2026, that figure moved into the 250-to-499 band, meaning the latest floor is higher than the previous ceiling.

More in Cybersecurity

Related articles

Cybersecurity

CyberStrike AI - Detection for Defenders: Scripting, Tempo and Residue

CyberStrike is a free, open-source AI agent for offensive security that pairs any LLM with 56+ tools, including a 40,000-line Windows post-exploitation module. A source-code reading of commit a5ed1e0 (v1.1.16) shows every Windows action runs through five fixed wrappers after a hard-coded five-process capability probe, leaving default cs-* and CyberStrike* artefacts. Includes Sigma, Suricata, YARA and KQL detections and a response playbook.

4 Nov 2026
Cybersecurity

Cybersecurity Checklist in the Agentic Hacking Era

Google Threat Intelligence Group and Mandiant say threat actors are now using agentic offensive frameworks such as ARTEX and PentAGI in real intrusions, moving from first stolen account to deep access in hours, sometimes 15 minutes. With CERT-In already warning of AI-assisted payment API hijacking at Indian NBFCs and fintechs, here is a checklist built on Google's five defence layers and CERT-In's controls.

2 Nov 2026
Cybersecurity

India's CERT Runs AI Cyberattack Drill for 55 Financial Institutions amid rising AI based attacks

India's Computer Emergency Response Team (CERT-In) held a full-day cybersecurity exercise and drill for the banking, financial services and insurance (BFSI) sector on 27 October 2026. More than 300 officials from over 55 institutions took part, mostly Regional Rural Banks and Cooperative Banks. The drill focused on detecting and countering AI-enabled attacks, including automated discovery and exploitation of vulnerabilities.

29 Oct 2026