OpenAI launches Trusted Access for Cyber initiative to bolster AI security

OpenAI has introduced a new initiative called Trusted Access for Cyber, designed to enhance digital defenses and manage the inherent risks associated with advanced artificial intelligence systems. This program aims to improve baseline protection for all users while providing controlled access to sophisticated cybersecurity capabilities for vetted defenders, according to a recent report by The Cyber Express.

The initiative leverages frontier AI models, including GPT-5.3-Codex, identified as OpenAI's most cyber-capable reasoning model to date. These advanced models can significantly accelerate vulnerability discovery, aid in faster remediation, and bolster resilience against targeted attacks, according to the company. However, their misuse poses serious risks. Trusted Access for Cyber seeks to unlock the defensive potential of these models while mitigating abuse.

OpenAI is also committing $10 million in API credits to support defensive cybersecurity work. The program allows individual users to verify their identity and enterprises to request trusted access for teams. An invite-only program is available for security researchers needing more permissive models. All users must adhere to OpenAI's usage policies.

Source: The Cyber Express