OpenAI Follows Anthropic in Limiting Access to Its Cyber-Focused Model

The article describes industry concern over AI models (Anthropic’s Mythos and OpenAI’s GPT‑5.4‑Cyber) that substantially improve automated vulnerability discovery and exploit development; vendors and regulators are limiting access via programs like Trusted Access for Cyber and Project Glasswing to reduce misuse, while warning that availability of such capabilities raises the likelihood of large-scale, automated attacks even as defenders gain faster patching and analysis tools.