Microsoft Sees Unicode Smuggling in Phishing Campaign
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.