Meta Platforms is making safety warnings for its AI assistant, Muse, more prominent after a security researcher found a significant vulnerability. Reported through Meta’s bug bounty program, the flaw could have allowed an attacker to access sensitive personal information, including emails and files stored in a user’s cloud-based virtual machine.
An internal incident report initially classified the vulnerability as medium-to-high severity on Meta’s internal scale. Meta has deployed updates to its warnings, noting that exploiting the flaw would typically require a user to interact with a prompt or click a warning. The incident highlights the security challenges of integrating AI agents more deeply into users’ digital lives.