AI Disclosure

What AI does here

• Writes plain-English explanations of findings (labeled “AI-generated explanation”).
• Fills in specifics for fix prompts built from our templates.
• Filters obvious false positives and classifies page text for some liability checks.
• Judges responses during AI-feature (LLM) security tests you enable.

What AI can't do

Models have no tools, network or database access. Their output can't start scans, open pull requests or send messages; it only becomes text a person reads.

Provider and data

We use Anthropic's Claude models under terms that prohibit training on our inputs. Only redacted data is sent. Scanned website content is treated as untrusted and the model is instructed never to follow instructions inside it.

Report a bad explanation

Every AI explanation has a “Report a bad explanation” link.