Meta’s AI support chatbot has been manipulated by hackers to infiltrate the Instagram accounts of famous individuals, Meta confirmed, saying the problem has now been resolved.

Compromised accounts ranged from Barak Obama’s White House account to beauty outlet Sephora.

According to reports from 404 Media, hackers asked Meta’s AI support chatbot to change the email addresses used by the targeted accounts.

Meta has installed AI support across its social sites – Facebook and Instagram – since March, giving the AI support bots the ability to reset passwords and account emails, as well as other account functions.

The methods used to manipulate the chatbots and gain account access were shockingly easy – screenshots from hacking groups and researchers reveal that hackers simply requested the chatbot change the email associated with targeted accounts, which then simply asked for verification code from said email in response.

Hackers were then able to use the chatbot to send a link to their email to change the targeted account’s password. In one video shared by a hacker, a VPN was used to change the hackers’ location and avoid some of Meta’s safeguards.

‘Manipulation’ seems like a strong word to discribe the simplistic and direct conversations hackers had with these chatbots which allowed them access to high profile accounts.

Recommended reading

• Tech Layoffs Jump 40% in Q1’26 as AI Fever Spikes
• Meta to Cut 10% of Workforce
• Oracle Announces “Significant” Layoffs As AI Pressures Squeeze
• Amazon Plans Largest Corporate Layoff Round in Decades

Meta said that the issue has now been resolved, and the firm is securing accounts impacted by the hack. It remains unclare how many accounts were affected in the operation.

The issue does raise concerns surrounding the increased reliance on generative AI systems and chatbots for secuirty and identity management.

Meta continues to lay off workers in favour of integrating AI across more and more of its systems and operations – should this instance give the tech titan pause on how dangerous AI systems can be if give high authority without a human-in-the-loop?