Latest
Virginia Democrat says DOJ wiretapped phone calls with state Senate leader4 House Democrats heading to Dallas for RNC counterprogrammingJudge declares mistrial in Lindsay Clancy murder case after jury fails to reach unanimous verdict – live updatesThe White House is making arcade games racistTrump threatens to halt trade with some countries unless the Federal Reserve lowers interest rates – US politics liveBlue Origin rocket explosion sent sound waves 1,000 miles across the US, report findsSix-week-old child in Pennsylvania dies from measles, coroner saysVirginia Democrat says DOJ wiretapped phone calls with state Senate leader4 House Democrats heading to Dallas for RNC counterprogrammingJudge declares mistrial in Lindsay Clancy murder case after jury fails to reach unanimous verdict – live updatesThe White House is making arcade games racistTrump threatens to halt trade with some countries unless the Federal Reserve lowers interest rates – US politics liveBlue Origin rocket explosion sent sound waves 1,000 miles across the US, report findsSix-week-old child in Pennsylvania dies from measles, coroner says
← Signal: Tech & AI

ASCII smuggling technique migrates from AI attacks to mass email spam evasion

Confirmed1 source · Sep 4, 2026

A method originally developed to hide malicious prompts from AI systems is now being weaponized by spammers to bypass email security filters.

ASCII smuggling technique migrates from AI attacks to mass email spam evasion
Image via Ars Technica

What happened

ASCII smuggling, a technique that uses invisible Unicode tag characters to encode text readable by computers but imperceptible to humans, has been adopted by spammers to evade email platform filters. Microsoft detected a massive spike in ASCII smuggling signatures beginning in early February 2024, jumping from approximately 21,000 per day to over 1.3 million within one day, reaching 2.5 million within four days. The wave of spam using this technique persisted for months before sharply declining in mid-May.

Context

ASCII smuggling exploits the same mechanism that made it effective for prompt injection attacks on AI systems—invisible encoding that machines process but humans do not perceive—to obfuscate spam keywords from automated detection systems. This application inverts the original intent but leverages identical technical properties, allowing malicious content to pass through security filters while appearing innocuous to end users. The technique demonstrates how defensive evasion methods can migrate across threat landscapes as adversaries adopt proven obfuscation approaches for different attack vectors.