AI in Russia Cuts Dangerous Emails in Half, Scaring Off Spammers
Improved algorithms and neural networks are making it increasingly difficult to bypass anti-spam systems.
The number of malicious emails in Yandex Mail fell by half in 2025 thanks to artificial intelligence, according to a blog post by Yandex 360.
Out of roughly 81 billion incoming emails processed over the year, nearly 4.8 billion messages were classified as malicious and blocked during processing. Another 9.8 billion emails were sent to the Spam folder by the SpamoOborona (SpamDefense) system. Last year, the number of blocked messages was almost twice as high. Common types of spam included fake prize and casino offers, phishing sites mimicking the Gosuslugi public services portal, and similar scams.
The overall share of spam also declined compared with the previous year, dropping from 25 percent to 18 percent. At the same time, scammers were noticeably less likely to attempt to harm email users. The development of artificial intelligence played a significant role in this shift.
Artificial Intelligence Takes Center Stage
Spam-fighting systems rely on algorithms and neural networks that check emails against around 10,000 criteria, ranging from the sender’s address to message structure and attachments.
This year, the already complex SpamoOborona technology was enhanced with logo recognition capabilities, known as computer vision anti-phishing, as well as optical character recognition for detecting text embedded in images.
Spam detection models are now retrained after each attempted attack, learning to identify previously unknown spam signals and factoring them into future blocking decisions. Neural networks adjust protection methods almost in real time by learning from scammers’ tactics.
In 2025, Yandex also introduced a rechecking mechanism called ReCheck. The tool can block or move potentially dangerous emails to spam even after they have already been delivered.
