Meta is making Muse AI security alerts more prominent after a bug-bounty researcher found a flaw that could expose users’ private emails and files. An attacker would need to trick a user into asking Muse to process a compromised link and manually approve a system prompt.
The Information reported, citing internal documents, that Meta initially rated the flaw SEV-2, its second-highest severity level. A Meta spokesperson said it was miscategorized and downgraded to SEV-3.
A separate Mac-related issue reported by an independent analyst involved local malware that could theoretically redirect Muse audio recordings to external servers and compromise account access. Meta said it posed a low operational risk and that a fix had been deployed.
