Skip to content

OpenAI releases GPT-6 Cyber to vetted security teams after rating it 'Critical' risk

OpenAI's new model for finding and exploiting software flaws went to a small group of approved customers on 5 November 2026. The company says it found 81 previously unknown high-severity bugs in Firefox.

Technology

By Comlumen Desk Compiled stories

3 min read

Compiled by the Comlumen Desk from the reporting and official sources listed at the end of this story, and written with AI assistance. It contains no original reporting by Comlumen. How we report

A corridor between racks of servers with blue lights
Rows of servers in a data centre. BalticServers.com / Wikimedia Commons, CC BY-SA 3.0

In this story
  1. What OpenAI says it can do
  2. Safeguards
  3. A dispute over data
  4. How this story was sourced

OpenAI released GPT-6 Cyber, an AI model built for computer-security work, on Thursday 5 November 2026. Access is limited for now to what Fortune described as "a small handful" of approved customers in Daybreak Red, its programme for offensive security work such as authorised penetration testing, in which teams attack systems with permission to find weaknesses before criminals do.

How this story was sourced.

4 sources from 4 organisations, 1 of them official or primary documents.

Official and primary documents 1

The body that made the decision, the paper, or the organisation’s own release.

  1. OpenAI Introducing GPT-6 Cyber openai.com

News organisations 3

Reports by other newsrooms. Where a story was carried by another outlet, both are named.

  1. Fortune OpenAI debuts new cybersecurity-focused AI model amid growing alarm about AI-driven cyber attacks fortune.com
  2. CSO Online OpenAI's GPT-6 Cyber ups the ante on vulnerability discovery csoonline.com
  3. CyberScoop OpenAI debuts GPT-6 Cyber, touting better performance and safety guardrails cyberscoop.com
The sources as the story lists them

Sources

Comlumen Desk Compiled stories
Stories under the Comlumen Desk byline are compiled from reporting by other news organisations and from official sources, and written with AI assistance. They contain no original reporting by Comlumen. Every source is listed and linked at the end of each story, and facts are attributed in the text to the organisation that reported them. How we report.

The story so far. More on Cybersecurity and AI.

  1. Technology Hackers tamper with fridges storing France's blood plasma; alarms prevent losses Attackers raised temperature settings at three cold-storage sites of France's national blood service overnight on 1 to 2 November 2026, through a refrigeration contractor. The service says no blood products were lost.
  2. Technology US tech stocks lose more than $2 trillion as investors weigh Democratic wins on AI The Nasdaq Composite fell 5.5% on 5 November 2026, its worst day since April 2025, as chipmakers led a sell-off driven by fears that a Democratic Congress will slow the data-centre build-out behind the AI boom.
  3. Technology AMD reports record $13.4bn quarter as data-centre sales nearly double AMD's revenue rose 45% in the third quarter of 2026 and it forecast about $17 billion for the fourth, but its shares fell 17% two days later in a broad chip sell-off.
More Technology
Made byStacklumen