Meta AI breaches external firm during security testing sandbox error

NPR | By Huo Jingnan,

Scott Simon

Published August 8, 2026 at 5:43 AM MDT

Meta announced one of its models had hacked another company during a security test. It becomes the third company to announced such a security breach.

Huo Jingnan

Huo Jingnan is a reporter for NPR.

See stories by Huo Jingnan

Scott Simon

Scott Simon is the host of Weekend Edition Saturday and is one of the hosts of NPR's morning news podcast Up First.

See stories by Scott Simon

Donate to the UPR Legacy Endowment

Honoring longtime UPR employees and station managers Kerry Bringhurst and Tom Williams, who are retiring this year, this endowment acknowledges the dedicated decades of service of Utah Public Radio employees over the station's 70+ years.

GIVE NOW