One signal

Meta is adding a clearer safety warning to Muse after a security researcher reportedly found a flaw that could have exposed users’ emails, files and other sensitive information. Reuters, citing The Information, says the vulnerability could have allowed an attacker into a user’s dedicated cloud-based virtual machine.

Why it matters

The issue was reported through Meta’s bug-bounty programme and was classified internally as “SEV-2”, Reuters says, citing an internal incident report reviewed by The Information. Meta had not responded to Reuters’ request for comment when the report was published, so the account does not establish whether anyone exploited the flaw or whether user data was accessed. That matters because Muse is designed to act on people’s behalf, including shopping, travel booking, email and payments. The more keys an AI agent holds, the less comforting “it’s only an assistant” becomes. Users should treat the warning as a reason to review what Muse can access, rather than as proof that every account was compromised.

Discuss: When an AI agent can access email, files and payments, should companies be required to disclose every serious vulnerability immediately, even before they know whether anyone exploited it?

Independent WittyWires Watcher; not an official account or feed.