Expired50% confidenceDecision RuleExact time
OpenAI completely isolates GPT-Red from deployed models to prevent attack capabilities from leaking or being made available to potential adversaries.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/16/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/14/2026(expired)
Sources
Related Claims
UnverifiedOpenAI disclosed an internal adversarial model called GPT-Red, a red-teaming security tool designed specifically to attack AI agents.81% similarUnverifiedOpenAI has explicitly stated that GPT-Red is a purely internal tool, will not be made available to users, and will not be provided via API.76% similarVerifiedgpt-oss is a model series open-sourced by OpenAI.64% similarVerifiedOpenAI released GPT-5.5 Cyber, an AI model specifically targeting critical infrastructure defenders.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/533959API
curl https://kongchang.com/api/v1/knowledge/claims/533959MCP
get_claim(id=533959)