Does GPT Red work in the EU?

Unverified. No explicit evidence found; the model is reportedly internal to OpenAI and not publicly offered in any region. Last verified 14 Nov 2026.

How much does GPT Red cost?

Pricing unknown. Internal model; no public access or pricing reported.

Where was GPT Red shown?

In 1 creator video by AI Search. Most recently by AI Search on 19 Jul 2026, at 30:52 in "Kimi K3, dancing waifus, robot UFC, song to MIDI, GPT Red, hoverboards: AI NEWS".

GPT Red

by OpenAI

OpenAI's internal red-teaming model trained via self-play to discover prompt-injection attacks; not publicly released.

UnverifiedResearch Language Models Agents & AutomationModel

Updated 14 Nov 2026EU status checked 14 Nov 2026Last shown 19 Jul 2026

GPT Red at a glance

What is GPT Red?: GPT Red is a model from OpenAI. OpenAI's internal red-teaming model trained via self-play to discover prompt-injection attacks; not publicly released.
Who makes GPT Red?: GPT Red is made by OpenAI.
Does GPT Red work in the EU?: Unverified. No explicit evidence found; the model is reportedly internal to OpenAI and not publicly offered in any region. Last verified 14 Nov 2026.
How much does GPT Red cost?: Pricing unknown. Internal model; no public access or pricing reported.
Where was GPT Red shown?: In 1 creator video by AI Search. Most recently by AI Search on 19 Jul 2026, at 30:52 in “Kimi K3, dancing waifus, robot UFC, song to MIDI, GPT Red, hoverboards: AI NEWS”.

Sources: creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.

About

GPT Red is an automated red-teaming model from OpenAI, described by a creator summary as trained through self-play to find prompt-injection attacks against AI systems. According to that secondhand account, it reached roughly 84% attack success, compared with about 13% for human red-teamers.

Its attacks were reportedly used as training data to harden GPT 5.6, which is said to have had about six times fewer failures than GPT 5.5 on these attacks. The figures come from a YouTube creator's summary and could not be checked against an official page.

GPT Red is reported to remain internal. It is not offered as a product, API or downloadable model, and no official documentation, paper or pricing was available.

Key features

Automated red-teaming via self-play
Targets prompt-injection attacks
Reported ~84% attack success vs ~13% for humans
Attacks used to train GPT 5.6 (reported ~6x fewer failures than 5.5)
Internal use only

1 moment

As seen on YouTube

Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.

AI Search19 Jul 2026

Kimi K3, dancing waifus, robot UFC, song to MIDI, GPT Red, hoverboards: AI NEWS

Internal automated red-teaming model trained via self-play to find prompt-injection attacks, reaching about 84% attack success versus 13% for humans. Its attacks were used to train GPT 5.6, which had six times fewer failures than 5.5; it stays internal.

30:52Open on YouTube