Threats · 52 days ago
Microsoft says a sustained phishing campaign has been using invisible Unicode tag characters to split financial lure words before email filters parse them, and telemetry from Defender for Office 365 showed the majority of those messages were still caught by layered protections. The technique was first popularized in AI prompt-injection research as ASCII smuggling, but here it was turned against traditional phishing defenses.
The trick is simple: the message looks normal to a person, but raw text handed to software contains hidden characters that break up words such as “funding,” so word-based parsing can miss the bait. That means a single Unicode-specific rule is not enough; the blind spot sits in the parsing pipeline, not in any one visible link or attachment.
For teams that rely on content-based filtering or message-parsing assistants, the exposure is whoever trusts a clean-looking lure as proof the text is clean. Microsoft’s telemetry also suggests layered controls help, but the reporting does not show that this obfuscation is limited to one campaign or one filter stack.
5 sources covering this story
Attackers conceal phishing lures using invisible Unicode characters
Threat actors have adopted the ASCII smuggling technique in phishing campaigns, using invisible Unicode characters to evade email security filters.
ASCII smuggling isn't just an AI security risk
Phishers find a new use for invisible Unicode tag characters
Once popular for attacking AI, ASCII smuggling is embraced by spammers
A once-overlooked block of unicode that's invisible to humans is gaining ever wider use.
Phishing Campaign Sends Millions of Emails Using Invisible Unicode to Evade Filters
A phishing campaign sent up to 2.37 million weekday emails using invisible Unicode tags to split financial lure words and bypass filters.
ASCII smuggling crosses over from AI prompt injection to phishing evasion | Microsoft Security Blog
Invisible Unicode characters popularized for hiding instructions from AI models are now being used to obfuscate words before email filters parse them.
Part of the PlainSec briefing for 2026-09-07