Anthropic and OpenAI now racing on purpose-built cybersecurity models

Anthropic released Mythos, a cyber-focused model, shortly before OpenAI expanded its Daybreak service with GPT-5.6-Cyber on August 10, 2026. The days-apart timing shows frontier AI providers competing directly on cybersecurity-specific models rather than treating security as a downstream application of general-purpose systems.