← Accueil actu | Salle du marché
2026-09-10 · Source：HotHardware
The technique behind AI prompt injection attacks has been repurposed to obfuscate spam keywords, Microsoft warns, combining with long-standing keyword smuggling to evade email detection filters at scale.
Microsoft is warning that a concealment technique first seen in attacks targeting AI systems has crossed over into ordinary email spam, appearing across millions of messages. The method, known as ASCII smuggling, is now being combined with an older, well-established trick — keyword smuggling — that spammers have long used to push unwanted content past detection filters. The overlap of the two techniques marks a notable escalation in the ongoing contest between spam operations and email defenses.
Keyword smuggling is hardly new. As Microsoft describes it, the practice involves obfuscating the keywords that would otherwise flag a message as spam, allowing it to bypass detection filters that scan for those terms. The tactic has been common practice for quite some time, and filter operators have continually adapted to it. What has changed, according to the company's warning, is that this traditional approach is now being enhanced by a new-school method borrowed from the AI threat landscape.
ASCII smuggling earned its reputation as a tool for AI prompt injection attacks, where it was originally deployed. In that context, the Unicode-based trick allowed hidden content to travel inside text in ways that were difficult for humans to see. Microsoft's warning is that the same technique has now found a second life in email, where it is being used to hide spam signals from the filters designed to catch them — at a scale measured in millions of messages.
For the security community, the development shows how a technique honed against AI systems does not stay confined to that arena. A method that emerged in prompt injection attacks is now compounding a problem as old as email filtering itself. Defenders already contending with keyword obfuscation must now reckon with concealment that operates at the Unicode level, a shift Microsoft has chosen to flag publicly as the technique spreads through inboxes at scale.
Mots-clés: Microsoft · ASCII Smuggling · Prompt Injection · Email Security
萬安算交所 AIXX · AIXX · Bourse compute vérifiée · Accéder à la salle