Updated
Updated · Ars Technica · Sep 4
Microsoft Detects 2.5 Million ASCII-Smuggling Spam Signatures in 4 Days
Updated
Updated · Ars Technica · Sep 4

Microsoft Detects 2.5 Million ASCII-Smuggling Spam Signatures in 4 Days

3 articles · Updated · Ars Technica · Sep 4

Summary

  • Microsoft Defender for Office saw ASCII-smuggling detections jump from about 21,000 a day to 1.3 million in early February, then hit 2.5 million within four days.
  • The technique hides spam keywords in 128 Unicode tag characters that computers can read but humans can barely see, letting messages slip past filters built to catch mass campaigns.
  • Microsoft said the same mechanism drew notice two years ago for concealing prompt-injection attacks on AI agents, but spammers have now repurposed it for email evasion.
  • The surge lasted for months before dropping sharply in mid-May, underscoring how AI-era obfuscation methods are spilling into broader cybercrime tactics.

Insights

How did a stealthy AI-hacking trick secretly infiltrate millions of everyday emails before Microsoft finally caught on?
Why did attackers suddenly abandon their massive invisible text campaign in mid-May despite successfully bypassing traditional keyword filters?