OpenAI GPT-5.5 and GPT-5.5-Cyber Release

OpenAI has released GPT-5.5-Cyber in limited preview for defenders of critical infrastructure and introduced the Trusted Access for Cyber (TAC) framework. This initiative aims to democratize AI-powered defense by providing verified security professionals with reduced classifier-based refusals for authorized cybersecurity workflows while maintaining strict safeguards against malicious activity.

Trusted Access for Cyber (TAC) Framework

The Trusted Access for Cyber framework is an identity and trust-based system that ensures enhanced cyber capabilities are available to verified defenders. The primary goal is to allow authorized users to perform defensive tasks without unnecessary model refusals, while still blocking requests that enable real-world harm, such as credential theft and malware deployment.

Access Levels and Use Cases

OpenAI defines three distinct levels of access to its models for cybersecurity work:

Access Level	Changes	Intended Use Cases
GPT-5.5 (Default)	Standard safeguards for general-purpose use	General-purpose, developer, and knowledge work
GPT-5.5 with TAC	Precise safeguards for verified defensive work	Secure code review, vulnerability triage, malware analysis, detection engineering, and patch validation
GPT-5.5-Cyber	Most permissive behavior for specialized workflows	Authorized red teaming, penetration testing, and controlled validation

Security Requirements for Access

To prevent unauthorized access to these permissive models, OpenAI requires phishing-resistant account security. Starting June 1, 2026, individual members of TAC must enable Advanced Account Security. Organizations may alternatively attest to having phishing-resistant authentication within their single sign-on (SSO) workflow.

GPT-5.5 vs. GPT-5.5-Cyber Capabilities

While GPT-5.5 is the primary model for most defenders, GPT-5.5-Cyber is a specialized version designed for higher-risk workflows.

• GPT-5.5: The smartest and most intuitive model for general-purpose knowledge and cybersecurity tasks, recommended as the starting point for most security workflows.
• GPT-5.5-Cyber: This model is not intended to significantly increase raw cyber capability beyond GPT-5.5; rather, it is trained to be more permissive. It is designed for specialized dual-use workflows, such as validating exploitability in controlled environments, where standard TAC access might still trigger refusals.

Ecosystem Integration and Partnerships

OpenAI is partnering with security vendors to create a "security flywheel" where AI capabilities are integrated into discovery, development, detection, and response layers.

Network and Security Providers

Providers use GPT-5.5 to reduce exposure through rule review, configuration analysis, and incident investigation. This allows for the deployment of WAF rules and edge mitigations before full remediation is complete. Cisco's SVP, Chief Security & Trust Officer Anthony Grieco, notes that these models act as a "force multiplier for defenders," increasing the velocity of operations.

Vulnerability Research and Patching

GPT-5.5 with TAC supports root cause tracing, mapping affected surfaces, and reviewing patches. For more specialized research requiring exploit proof-of-concepts for coordinated disclosure, GPT-5.5-Cyber is utilized by a smaller set of partners.

Detection and Monitoring

Partners like SentinelOne use GPT-5.5 to help analysts connect telemetry signals and draft detections more quickly, moving from disclosure to investigation faster.

Software Supply Chain Security

Tools from partners such as Snyk, Gen Digital, Semgrep, and Socket use GPT-5.5 with TAC to inspect dependency changes and identify suspicious package behavior early in the development cycle. Snyk's Chief Innovation Officer Manoj Nair describes the access as a "strategic necessity" to counter attackers weaponizing frontier models.

Codex Security for Open Source

OpenAI is also investing in open-source maintainers through Codex Security, which helps identify and remediate vulnerabilities by building codebase-specific threat models and proposing patches.

• Codex for Open Source: Selected maintainers of critical projects receive conditional access to Codex Security and API credits.
• Codex Security Plugin: A plugin is available to bring security workflows—including attack-path analysis and verified fixes—directly into Codex interfaces (app or CLI).

Future Outlook

OpenAI intends to broaden access to these capabilities as identity verification, approved-use scoping, and misuse monitoring improve. During alpha testing, GPT-5.5-Cyber has already been used to scale automated red-teaming of critical systems and validate high-severity vulnerabilities.