Mythos & GPT-Cyber: AI's Dual Impact on Cybersecurity

New artificial intelligence models are changing how cybersecurity works. Anthropic announced Mythos Preview, an AI that can find and fix cybersecurity vulnerabilities. It has already identified thousands of undiscovered zero-days. OpenAI also unveiled GPT-5.4-Cyber, an updated AI specifically for cybersecurity problems. They later released GPT-5.5-Cyber. Both companies are currently limiting access to these powerful tools. Mythos Preview is available to partners in Anthropic’s Project Glasswing, including major tech companies. OpenAI's GPT-Cyber is for vetted members of its Trusted Access for Cyber program. The concern is that these AI tools could be misused by cybercriminals, who are already using AI for attacks. Experts warn that these frontier AI models could uncover so many vulnerabilities that organizations might struggle to patch them all. This could create a "vulnpocalypse," where criminals exploit newly disclosed vulnerabilities before companies can fix them. This matters because it highlights a growing challenge for protecting digital systems.