Once popular for attacking AI, ASCII smuggling is embraced by spammers

SummaryAI generated

ASCII steganography embeds malicious instructions into email content using Unicode tags. These characters are invisible to humans but can be recognized and parsed by AI. This technology was originally used to evade AI attacks, but now spam senders use it to bypass email filtering systems, utilizing the highly similar coding characteristics of standard ASCII for covert transmission.

Related eventsRELATED EVENTS

2026-09-06 05:17OpenAI responds to attacks on its agents on German websites: Will completely reform the disclosure mechanism for “AI model attacks” incidents
2026-09-08 08:00Split-LLM training gradient leakage privacy: Cloud-based systems use zero-gradient mode to accurately reconstruct real data rows
2026-09-08 19:30LG smart TVs will scan the home network and record audio through the microphone in standby mode

Key entitiesKEY ENTITIES

ASCII smuggling

SignalsSIGNALS

Keyword heat

ASCII smuggling1

All reports (1)SOURCES

↗

A Ars Technica en 2026-09-05 01:18

ASCII smuggling is a technique once used to hide AI attack instructions, now adopted by spam senders to bypass email platform filtering mechanisms. This technique embeds malicious instructions into content using Unicode tags (such as U+E0041 for “A” and U+E0061 for “a”). Although these characters can be recognized by computers, they are almost invisible to humans. Originally used to evade AI prompt injection attacks, it is now used to bypass email filtering systems. This technique relies on 128 ASCII tags, which are highly similar to standard ASCII, with only differences in readability. This makes malicious content visually invisible yet still detectable and executable by AI models.