AegisAI Releases Proteus Email Security Model

AegisAI announced in a press release the release of Proteus, a foundation model that investigates incoming emails before they reach inboxes. The system examines message content, headers, sender and recipient history, link and attachment data, and threat intelligence.

AegisAI says Proteus detected more than 99% of 3,020 verified phishing attacks collected from one week of customer traffic. A leading closed source general model detected 81%, while an open weight model detected 80%. Proteus returned verdicts up to 7.2 times faster at the median, according to the company.

The benchmark measured missed detections but did not include false positive results. The general models received only the email and a classification instruction, while Proteus received additional evidence collected by AegisAI. The company says Proteus will undergo continuous retraining and testing across nine attack categories.