AI Lab
Interactive tools from CyberWatch Daily — small, focused demos that show how LLMs can assist security workflows. Not a substitute for professional review.
Phishing email classifier
Paste an email body to receive a risk score, verdict, and the signals behind the assessment.
Architecture notes
Scored by a PHP rule engine running on the same host as the site: lookalike-domain normalisation, urgency and credential-request patterns, link reputation heuristics, and header mismatches, each carrying a weight that adds up to the risk score. Rate-limited per IP. Text never leaves the server, and the report states what the check could not see.
UI copy risk checker
Paste permission or consent dialog copy to get clarity feedback and a dark-pattern risk score.
Architecture notes
Same engine with a consent-UX rule set: pre-selected choices, steering labels, unnamed third parties, missing decline paths, and confirmshaming each add weight, and the endpoint returns a rewrite scaffold. Text-only, so it cannot judge layout, contrast, or button size.