Hackers reportedly found a way to use Meta’s AI support chatbot to gain access to Instagram accounts, including several high-profile profiles, according to a report from 404 Media. The method was strikingly simple: attackers asked the chatbot to help change the email address tied to a target account.

The reported abuse came to light alongside a string of account takeovers involving notable Instagram profiles, including the Barack Obama White House account, the Chief Master Sergeant of Space Force account, and Sephora’s account. The report says screenshots and videos shared in Telegram groups among security researchers and hackers showed how the process worked.

In one example described by 404 Media, a hacker initiated a conversation with Meta’s support bot and requested that a target username be linked to a new email address. The message included instructions suggesting the attacker would supply a verification code, a tactic that appears to have been enough to move the account-change process forward.

The incident highlights a broader concern about relying on automated support systems for sensitive account functions. Meta had recently expanded its AI support tools across Facebook and Instagram, saying in March that the assistant would be able to help with account recovery tasks, including password resets and other maintenance functions. The company described the feature as providing not only suggestions but direct solutions for account security and recovery.

That capability now appears to have introduced a serious risk. According to the report, users who lost access to their accounts said they had no way to escalate their cases to a human support representative. If true, that would leave victims dependent on the same automated system that attackers allegedly manipulated.

The vulnerability also raises questions about how much trust should be placed in AI systems handling identity verification and recovery. Account recovery tools often deal with the most sensitive parts of a user’s digital life, including email addresses, passwords and recovery settings. A failure in that process can lock legitimate users out while giving attackers an opening to take control.

Telegram groups circulating the screenshots and videos suggest the exploit was widely discussed among both security researchers and hackers before the issue became public. The material described in the report points to a flaw that did not depend on advanced technical intrusion, but rather on persuading the chatbot to perform a critical administrative action.

Meta has since fixed the security flaw, according to the source report. The company did not, in the material provided, offer a detailed public explanation of how the problem occurred or how the fix works. Still, the episode serves as a warning that AI-powered support tools can create new attack surfaces when they are given access to account control features.

For companies rolling out AI agents to handle customer support, the case underscores a difficult balance. Automation can make recovery faster and cheaper, but if safeguards are weak, the same tools can become a path into accounts that should be protected. In this case, the risk was not theoretical. Hackers reportedly tested the system and, at least for a time, it worked.