<![CDATA[⚪️ Font rendering tricks AI assistants into approving malicious commands]]>⚪️ Font rendering tricks AI assistants into approving malicious commands

🗨️ Researchers from LayerX have developed a proof-of-concept attack that makes it possible to hide malicious commands from AI assistants. The attack is based on a discrepancy between what the AI sees in the page’s HTML code and what is actually…

🔗 https://hackmag.com/news/poisoned-typeface?utm_source=mastodon&utm_medium=social&utm_campaign=repost_hackmag_to_socials

]]>https://board.circlewithadot.net/topic/4933a14c-a101-4f63-a617-1b52afaff343/font-rendering-tricks-ai-assistants-into-approving-malicious-commandsRSS for NodeSat, 18 Apr 2026 02:18:36 GMTFri, 20 Mar 2026 08:30:03 GMT60