Meta says it has resolved a problem that allowed hackers to trick its AI assistant into granting entry to different customers’ accounts, together with high-profile people, in line with a number of media reviews.
404 Media and The Guardian say hackers used the tactic to focus on Barack Obama’s White Home account, magnificence retailer Sephora and the U.S. House Drive chief grasp sergeant, John Bentivegna.
Common customers additionally reported comparable takeovers on X, even sharing screen-recorded movies of conversations with Meta’s AI chatbot detailing how the hacks have been executed.
One video posted on X seems to point out a hacker asking Meta’s AI assistant to hyperlink their account to a brand new e-mail handle, bypassing two-factor authentication.
Get every day Nationwide information
Get every day Canada information delivered to your inbox so you may by no means miss the day’s prime tales.
The bot responds {that a} verification code has been despatched to the brand new e-mail handle and asks the hacker to enter it within the chat. As soon as the hacker does so, an possibility seems permitting them to reset the hacked consumer’s password.
“This challenge has been resolved and we’re securing impacted accounts,” Meta stated in an announcement to World Information on Tuesday.
The breach comes as Meta undergoes an operational shift that can see human roles transferred to AI and a rise in the usage of tech-forward options throughout its platforms, together with buyer help on Instagram and Fb.
The corporate’s chatbot buyer help function, launched final yr, will proceed to be up to date over time, in line with a March replace from the tech big.
Meta stated within the launch that its AI chatbot is about up for customers to report scams, impersonation accounts or problematic content material and to handle privateness, reset passwords, and replace profile settings.
Final yr, Meta reported “optimistic outcomes” from modifications to the service, which it stated lowered errors and targeted on regulating unlawful and extreme content material throughout all Meta platforms, together with terrorism, little one exploitation, medication, fraud and scams.
Meta CEO Mark Zuckerberg arrives for a landmark trial over whether or not social media platforms intentionally addict and hurt youngsters, Feb. 18, 2026, in Los Angeles.
AP Picture/Ryan Solar
The AI methods will “finally be capable of tackle work that’s higher suited to know-how, like repetitive opinions of graphic content material or areas the place adversarial actors are always altering their techniques, resembling with illicit drug gross sales or scams,” the March launch famous, earlier than including that Meta will “nonetheless have folks who assessment content material.”
In January, lots of of Meta workers have been laid off in its Actuality Labs division because the tech big shifted its focus from metaverse merchandise to AI. In late Might, Reuters reported that CEO Mark Zuckerberg stated in a company-wide memo that the corporate was not planning any additional layoffs for 2026.
He made the announcement on the identical day the Fb proprietor carried out a huge restructuring of the corporate, shedding 10 per cent of its workforce globally and transferring 7,000 different workers to new initiatives associated to AI workflows.
— With information from Reuters
© 2026 World Information, a division of Corus Leisure Inc.
