Meta patches Muse exploit that let attackers control the AI agent

The zero-day exploit required section entree to the user’s device, but gave imaginable attackers entree to Muse accounts. | Image: The Verge

Meta has issued a spot for its Muse macOS app pursuing the find of a zero-day vulnerability that could let idiosyncratic to instrumentality power of the AI agent. The bug recovered by security researcher Patrick Wardle utilized an undocumented Muse mounting that enabled imaginable attackers moving section codification to redirect transcription processing from Meta's servers to their ain endpoint, Ars Technica reports, giving the attacker entree to the Muse account.

Several plan decisions reportedly enabled this flaw, including having Muse dictation hap successful the unreality alternatively of on-device, and allowing immoderate app to power each of Muse's undocumented settings. Pr …

Read the afloat communicative astatine The Verge.