Definition
An AI model built and trained specifically for security tasks — like finding vulnerabilities or writing patches — rather than a general-purpose chatbot repurposed for the job. Microsoft, Google, and Cisco have each recently shipped their own versions, arguing they're faster, cheaper, and don't require sending sensitive source code to a general external AI service.
Why it matters
This signals a maturing market for security-specific AI tools; boards evaluating AI security vendors should ask whether a product uses a purpose-built model versus a general-purpose one repurposed with a thin wrapper.