Israeli Cybersecurity Experts Uncover AI-Powered Dark Sourcery Scam

Israeli cybersecurity experts have identified a sophisticated new scam dubbed "Dark Sourcery," which exploits artificial intelligence to trick users into contacting fraudulent support services. Attackers are "poisoning" data sources used by popular AI models, such as ChatGPT and Google Gemini, by creating fake support pages, reviews, and articles optimized for search engines. When AI models scan these compromised sources to answer user queries, they incorporate the false information, presenting malicious contact details as legitimate.

The campaign has targeted 374 well-known brands, including major airlines like Lufthansa, United Airlines, and Emirates, booking platforms like Airbnb and TripAdvisor, and financial institutions such as Chase, Citi, and Wells Fargo. The method does not involve hacking the AI models themselves but rather manipulating the data they access. This "search poisoning" technique leads AI to provide fake phone numbers or phishing links when users seek official company contacts.

This poses significant risks, as users attempting to resolve issues, like changing a flight ticket via an AI chatbot, might be directed to fake call centers. Scammers then impersonate support staff to steal banking information or charge exorbitant fees for services. When alerted, tech giants have shown mixed responses. Google stated such misinformation falls outside its vulnerability disclosure program, while OpenAI closed a report due to the difficulty in consistently reproducing the error, as AI responses are dynamic.

Cybersecurity professionals strongly advise users to always verify support contact information directly on official company websites, rather than relying solely on AI-generated recommendations. This warning follows a previous report by "Cursor" about scams targeting Israelis before the Sukkot holiday, exploiting unexpected avenues.