Executive Summary
Microsoft researchers have identified a high-volume phishing campaign that repurposes 'ASCII Smuggling'—a technique originally used in AI prompt injection—to bypass traditional email security filters. Starting in February 2026, the campaign utilized invisible Unicode tag characters (U+E0000-U+E007F) to fracture high-signal financial keywords like 'funding,' ensuring the words remain readable to humans while appearing as non-contiguous strings to automated detection systems. This shift demonstrates how adversarial techniques designed for AI systems are rapidly migrating into the broader cybercrime ecosystem.
The campaign primarily targets financial interests using SBA-themed lures, business loans, and lines of credit. It operates with a disciplined weekday-only schedule, peaking at over 2.3 million messages daily. The traffic originates from legitimate email marketing infrastructure, specifically ActiveCampaign, which complicates reputation-based filtering. While the primary goal is filter evasion rather than instruction smuggling, the adoption of this technique highlights a significant evolution in social engineering and obfuscation tradecraft.
For defenders, this represents both a challenge and an opportunity. While the technique can defeat naive keyword matchers and tokenizers, the presence of these deprecated Unicode characters is highly anomalous in legitimate business communication. Implementing normalization layers that strip or fold these characters before evaluation is critical for maintaining the efficacy of existing security controls.