As attacker techniques continue to evolve, we’ve introduced three upgrades that strengthen credential-phishing protection, expand detection coverage, and significantly reduce false positives across the platform.
Enhanced Credential-Phishing Protection
Credential-theft attacks increasingly rely on hidden redirects, brand impersonation, and file-sharing lures. Improvements to redirect analysis and new detectors for modern phishing techniques, supported by a high-capacity model trained on 400 million messages, have reduced missed credential-phishing attacks by 38% while lowering false positives by 30%.
Improved Accuracy for User-Reported Phishing
Enhancements to the AI Security Mailbox labeling and training pipeline deliver clearer, more consistent classification of employee-reported messages. These updates have reduced false positives by approximately 66% and further minimized rare cases where real threats are incorrectly classified as safe.
Expanded Detection Coverage
We’ve expanded our core and multimodal detection models to increase overall detection coverage by approximately 10%, improving identification of subtle and emerging attacks while preserving all existing protections and workflows.
These enhancements reflect our ongoing investment in people and AI-enabled systems, expanding protection against evolving attacks while keeping operations efficient and burden-free for security teams.