With the addition of the BERT LLM enhancement, Abnormal's detection models can more easily determine if two emails are similar and are part of the same polymorphic email campaign targeting an organization. Additionally, these pre-trained BERT LLMs give Abnormal the ability to understand content and text and the intention of a possible attacker in a highly scalable manner.

Here's an example of the differences a word's meaning can have depending on where it sits within a sentence.

Because BERT can understand the context of word placement in text, it will return different embeddings as vectors that encapsulate the meaning of the word to detect similar spam campaigns.