A vulnerability discovered in Meta’s AI-powered support system once again highlights the importance of security auditing for artificial intelligence solutions

According to information released by Malwarebytes, a U.S.-based company specializing in cybersecurity, a security vulnerability was discovered in Meta’s AI-powered support service used on Instagram to assist users with account-related issues. The report states that attackers were able to exploit a weakness in the functionality of the AI support bot to compromise several high-profile Instagram accounts.

The integration of artificial intelligence technologies into critical information systems and user account management creates new opportunities while also introducing additional cybersecurity risks. The incident serves as another example demonstrating that these risks are real and should be carefully considered when deploying AI-driven solutions.

The report notes that attackers abused the functionality of Meta’s AI support bot to add new email addresses to targeted user accounts and subsequently gain access to account recovery mechanisms. Notably, the attack did not involve traditional cyberattack methods, malware or stolen authentication credentials. Instead, the attack scenario was directly based on exploiting weaknesses in the decision-making and operational execution logic of the artificial intelligence system.

According to media reports, the cyber incident targeted and reportedly compromised several high-profile Instagram accounts, including the Obama-era White House Instagram account, Sephora’s account and an account belonging to a senior official of the U.S. Space Force. The incident demonstrates that granting AI-powered services sensitive privileges such as user account management, authentication processes and modification of access rights must be accompanied by additional security control mechanisms.

The incident once again brings to the forefront the importance of conducting independent security assessments, functional audits, penetration testing and risk analyses before deploying artificial intelligence solutions. Practice shows that an AI system which functions correctly from an operational perspective may create new attack surfaces for organizations if it is not sufficiently assessed from a security perspective. It should be particularly noted that, when integrating artificial intelligence systems with user accounts, identification and access control processes, the principle of least privilege must be applied, the execution of critical operations must be protected through multi-stage approval mechanisms and decisions made by AI must be supported by audit and continuous monitoring mechanisms. Otherwise, solutions developed to provide services to users may be exploited by cybercriminals and turned into real tools for account compromise, data leaks and other security incidents.

For information, according to the relevant Order of the President of the Republic of Azerbaijan on the approval of the “Artificial Intelligence Strategy of the Republic of Azerbaijan for 2025–2028”, the Special Communication and Information Security State Service of the Republic of Azerbaijan has been designated as the authority responsible for analyzing information security risks in areas where artificial intelligence is applied in state agencies (institutions), as well as taking the necessary measures to ensure information security related to artificial intelligence in state agencies (institutions).