OpenAI has announced the release of GPT 5.4 Cyber, a specialized version of its latest flagship model GPT 5.4, designed to support defensive cybersecurity operations. The announcement comes shortly after Anthropic introduced its frontier model Mythos, reflecting a growing focus among artificial intelligence companies on security oriented capabilities. GPT 5.4 Cyber is intended to assist professionals responsible for protecting systems, data, and users by enabling faster identification and remediation of vulnerabilities within digital infrastructure.
Alongside the model launch, OpenAI revealed that it is expanding its Trusted Access for Cyber program, extending availability to thousands of verified individual security professionals as well as hundreds of teams tasked with safeguarding critical software systems. The initiative is aimed at ensuring that advanced AI tools are accessible to legitimate defenders while maintaining strict controls to reduce the risk of misuse. OpenAI emphasized that artificial intelligence technologies are inherently dual use, meaning they can be applied for both protective and harmful purposes if not carefully managed.
The company acknowledged ongoing concerns that adversaries could potentially adapt defensive AI models to uncover and exploit vulnerabilities before they are patched, creating significant risks for widely used software. In response, OpenAI stated that its approach focuses on balancing broader access with strengthened safeguards, including protections against jailbreak attempts and adversarial prompt manipulation. This controlled and phased rollout strategy is designed to support responsible usage at scale while maintaining oversight as model capabilities continue to advance.
OpenAI also highlighted progress made through its Codex Security initiative, an AI driven application security agent developed to identify, validate, and propose fixes for software vulnerabilities. According to the company, the tool has already contributed to resolving more than 3000 critical and high severity vulnerabilities. This reflects a shift toward integrating AI directly into development workflows, allowing developers to receive immediate and actionable feedback during the coding process rather than relying solely on periodic audits or static vulnerability assessments.
The broader industry context underscores an increasing emphasis on using advanced AI systems to strengthen cybersecurity defenses. Anthropic’s Mythos model, which is being deployed under Project Glasswing in a controlled environment, has reportedly identified thousands of vulnerabilities across operating systems, web browsers, and other software platforms. OpenAI indicated that building a resilient ecosystem depends on continuously detecting and addressing security issues as software is created, with AI playing an increasingly central role in enabling proactive and ongoing risk reduction across the development lifecycle.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
