Meta AI Vulnerability Allegedly Enables Instagram Password Resets

Instagram is facing scrutiny after a critical vulnerability in its Meta AI-powered support system allegedly allowed attackers to take over user accounts by abusing the password recovery process.

The tool, designed to help users regain access to locked accounts, could be tricked into sending password reset codes to unauthorized individuals without proper identity verification.

The vulnerability did not involve a direct breach of Meta’s infrastructure. Instead, it existed within the AI assistant’s logic. Attackers reportedly initiated conversations with the chatbot and crafted prompts that convinced it to forward password reset links or codes.

Because the system lacked strong authentication checks and rate-limiting controls, anyone with knowledge of a target’s username could potentially trigger the process.

Meta confirmed that its backend systems were not compromised. However, the incident highlights a growing security concern around AI-driven support tools. In this case, the chatbot acted on requests without verifying whether the requester was the legitimate account owner.

According to the report, Security researchers ZachXBT and Dark Web Informer were among the first to publicly report the issue, revealing that threat actors had discovered a way to manipulate Instagram’s AI chatbot.

This type of vulnerability is particularly dangerous because it bypasses traditional security layers. Rather than exploiting code injection or server weaknesses, attackers leveraged social engineering techniques against the AI model itself.

This reflects a broader trend where attackers increasingly target automated systems that lack contextual judgment and strict enforcement policies.

Meta AI Vulnerability

Threat actors focused on premium Instagram handles, especially short and rare usernames that carry significant resale value in underground markets. Accounts such as @hey and @jowo were reportedly among the targets.

These accounts are often traded for large sums, sometimes exceeding hundreds of thousands of dollars. According to Dark Web Informer, stolen accounts were quickly listed and sold through private Telegram channels.

The speed of these transactions suggests a well-organized ecosystem where compromised accounts are monetized almost immediately.

This “account takeover as a service” model has been growing in recent years, with threat actors specializing in identifying, hijacking, and reselling valuable digital identities.

Meta responded to the reports by issuing a fix late Friday. In its official statement, the company said it had resolved an issue that allowed external parties to request password reset emails for some users. It emphasized that there was no system breach and that user accounts remain secure.

Despite this assurance, the incident has raised questions about how much access AI tools should have when handling sensitive operations like account recovery. The lack of strict validation controls in this case exposed a critical attack surface.

The vulnerability underscores the risks of integrating AI into security-sensitive workflows without robust safeguards. AI systems, while efficient, can be manipulated if they do not enforce strict authentication and context-aware decision-making.

Notably, accounts protected with two-factor authentication were not affected by this attack. This indicates that traditional security controls still play a crucial role in mitigating emerging threats.

The incident serves as a warning for both platforms and users. As AI becomes more deeply embedded in account management systems, attackers are likely to continue probing these tools for weaknesses.

Stronger validation mechanisms, improved rate limiting, and better AI behavior constraints will be essential to prevent similar abuses in the future.

Follow us on Google News, LinkedIn, and X to Get Instant Updates and Set GBH as a Preferred Source in Google.