DigiNews

Summary

Ars Technica reports on ASCII smuggling, a Unicode-based technique that hides malicious prompts inside text to evade AI and spam filters. The method, once discussed as an AI prompt-injection concern, has been adopted by spammers to defeat ML/NLP detectors, with Microsoft Defender for Office documenting a large spike in detections earlier in 2026. The article also explains how standard filters can miss such content and notes guidance from Microsoft on improving detectors.