After decades of establishing itself as one of the worst guardians of personal data, Meta raised eyebrows this year by encouraging users to hand over their email logins and credit card information to its new AI agent, Muse.
The tool, a privacy nightmare anthropomorphized by a furry apparition, can crawl through years of personal instant messages and even glean highly sensitive information by accessing entire email inboxes, from long-forgotten exchanges with exes to passport numbers.
To Meta’s former security chief Alex Stamos, it’s a “totally insane” idea to hand over some of this data, as Business Insider reports. During a recent appearance on Alex Kantrowitz’s “Big Technology Podcast,” Stamos discussed how some users are going as far as to hand over their password vaults, effectively giving Muse control over their entire digital lives.
It’s not just Muse hoping to cash in on the trend. Viral personal AI assistant startup Instinct, which was founded by a 23-year-old research scientist months ago and raised $1 billion in a September funding round, is similarly developing agents that can autonomously complete day-to-day tasks.
As a result, the financial world has a major headache on its hands as it tries to separate out legitimate users — and their AI bots — from suspicious activity. Worse yet, the way the AI agents go about doing their work on behalf of their human overlords requires them to break security measures by definition.
“They’re not stolen credentials, but are clearly bots doing interaction,” Stamos said during his podcast appearance. “And one of the reasons Instinct has a ridiculous valuation from my perspective, but one of the things they did that made them popular is they actually figured out ways to defeat a bunch of the automated bot detection mechanisms from these companies.”
Meta maintains on its website that Muse is safe because it runs on a virtual machine that is “isolated so no one else’s agent can access it.” However, “no automated system is perfect,” the company admitted, advising users to “review Muse’s activity and only grant the access you’re comfortable with.”
As such, security researchers have already discovered vulnerabilities that could give hackers access to sensitive personal data, as The Information reported in September, causing Meta to bolster its defenses. Then, last month, researchers once again identified other ways attackers can expose sensitive information when users share message exchanges through web links, per Ars Technica.
Then there was tech YouTuber Matt Robb, who claimed that Muse was giving out his address to strangers and even sent them to his home via his Facebook Marketplace account.
In short, Stamos’ calls for caution are entirely warranted, and should have anybody think twice about handing Meta the keys. And that’s not to mention the company’s abysmal track record when it comes to safekeeping personal data.
More on Muse: Meta’s Muse Is Spying on All Your Friends and Family
The post Former Meta Exec Shreds the Company’s Muse AI Agent as “Insane” Privacy Risk appeared first on Futurism.
