Meta discovered Muse flaws that could expose sensitive company data
Technology
In the weeks before launching its new AI assistant Muse, Meta discovered some pretty serious security flaws that could have let users peek at sensitive company data.
Meta started working on the issue on August 27, tied to a Linux exploit from July, sent Meta's teams into overdrive to lock things down.
Researcher found post-launch Muse control flaw
Starting August 27, Meta limited Muse's access to company systems and the internet while running extra tests and security audits.
Even after launch, one researcher found another vulnerability that could allow apps and terminal commands to control a user's Muse, showing just how tricky it is to keep advanced AI systems secure.