Why legacy filters fail and MxGuard doesn't

Rule-based and Bayesian filters were built for an era of obvious spam. They look for bad grammar, known-bad keywords and blocklisted senders.

Today's most damaging attacks look nothing like that. CEO impersonation, invoice redirection, gift-card fraud, lookalike domains and AI-written phishing with perfect grammar carry no payload and no obvious tell — they are short, clean, plausible messages. That is precisely what sails past rule-based and Bayesian filters.

MxGuard's LLM layer reads these borderline messages for intent and context, the way a trained person would, rather than matching them against a rulebook. That is why MxGuard catches the targeted attacks that legacy gateways miss, while keeping false positives on genuine mail extremely low.