MODEL

GPT-6 Cyber

modeltopic-noteopenaisecurity

Overview

GPT-6 Cyber is OpenAI‘s November 2026 purpose-trained frontier cybersecurity model — a specialised variant of GPT-6 Astra built for authorised vulnerability research, exploit development, and penetration testing. It is OpenAI’s first “Critical”-rated dual-use model under the Preparedness Framework (prior Daybreak models were rated “High”) and is restricted at launch to approved customers inside OpenAI’s Daybreak Red program, not GA. Successor in the Daybreak Red slot to GPT-5.6-Cyber.

Timeline

  • 2026-11-08-AI-Digest — OpenAI released GPT-6 Cyber on 2026-11-05 as a specialised variant of GPT-6 Astra for authorised vulnerability research, exploit development, and penetration testing, restricted at launch to approved customers inside OpenAI‘s Daybreak Red program. The Verge reports GPT-6 Cyber outperforms GPT-6 Astra on cyber benchmarks and goes out of bounds in less than 1% of graded runs — down from ~5.51% for GPT-5.6-Cyber — attributed to OpenAI’s own system card rather than independent third-party benchmarking. It is OpenAI’s fourth cybersecurity-focused model of 2026 and the first to carry the Preparedness Framework’s “Critical” cyber rating (prior Daybreak models were rated “High”). Reframe worth carrying: the newsworthy move isn’t the capability bump, it’s the ratings-tier flip — a “Critical”-tier model that can both find and exploit vulnerabilities is the highest-capability dual-use AI OpenAI has shipped, and the gated Daybreak Red rollout is a visible pivot away from broad distribution for frontier cyber capability. Pair against the Anthropic Responsible Scaling Policy’s ASL-3/ASL-4 tier structure — different taxonomies, converging operational answer. Log against MOC - Agent Security, MOC - Major Companies.

  • 2026-11-14-AI-Digest — GPT-6 Cyber is the model powering both halves of the Daybreak pair shipped today — Red Team (early access) for agentic web-app vulnerability discovery, bug-bounty report verification and patch testing, and Defense Factory for codebase-scanning + PR-with-fixes-and-tests shipping. The Nov 5–6 public release context (csoonline / thenextweb / bleepingcomputer cross-coverage) hardens into named-product carries today rather than a fresh launch beat. Log against MOC - Agent Security, MOC - Major Companies.

Key Developments

  1. First “Critical”-Tier Dual-Use Model Under the Preparedness Framework (November 2026): GPT-6 Cyber carries OpenAI’s first Preparedness Framework “Critical” cyber rating — a ratings-tier flip from the “High”-rated GPT-5.6-Cyber and prior Daybreak models. Shipped gated to Daybreak Red (approved customers only), not GA.
  2. Benchmark Delta vs GPT-5.6-Cyber (November 2026): Reported less-than-1% out-of-bounds rate on graded runs vs ~5.51% for GPT-5.6-Cyber, per OpenAI’s own system card (not third-party benchmarking). Capability-bump beat sits alongside the ratings-tier-flip beat but the latter is the load-bearing story.
  3. Fourth OpenAI Cyber Model of 2026: Follows GPT-5.4-Cyber, GPT-5.5 Cyber, and GPT-5.6-Cyber inside a single calendar year — OpenAI’s cybersecurity-model release cadence remains the highest of any US frontier lab. Pairs with Anthropic‘s RSP ASL-3/ASL-4 tier structure as a different taxonomy converging on gated-rollout as the operational answer.

See also: OpenAI, Daybreak, GPT-5.6-Cyber, GPT-5.5 Cyber, GPT-5.4-Cyber, GPT-6, Astra, MOC - Agent Security, MOC - Major Companies.