Threat · curated 3 Oct 2026
Microsoft built a prompt injection detector. Then it...
First reported daily.dev
Coverage timeline
Single-source incident — first reported, latest, and curated coincide.
Why it matters
ASCII smuggling via invisible Unicode tag characters lets attackers hide prompt injections in untrusted text that reaches LLMs and AI agents, bypassing normalization and keyword filters that defenders typically rely on.
Microsoft Defender for Office 365 detected a large-scale phishing campaign using invisible Unicode tag characters (U+E0000–U+E007F), an 'ASCII Smuggling' technique, with hunting-signature hits jumping from ~21,000 to over 2.3 million in three days. The same technique was previously demonstrated against LLMs to smuggle hidden prompt injections, and it threatens AI agent pipelines because standard Unicode normalization (NFC/NFD) does not strip these characters and tokenizers handle them inconsistently.