ChatGPT and Gemini can Hack their Users to Redirect to Fraud Sites

Artificial intelligence systems are located in the center of our daily lives, but this development brings new security threats. Recent research reveals popular chatbots like ChatGPT and Google Gemini can be directed to fake support services and fraud centers with the attacks of hackers.

How does Saldurgans do this job?

• Data poisoning: Hackers, adding fake web pages and scam links to educational data or indexing systems of AI models, enable models to evaluate this information as reliable source.

• Prompt injection techniques: By triggering chatbots to give certain responses through built-in commands, users lead to malicious sites.

• Social engineering: Saldurganlar is creating realistic-looking support services and help center pages and train chatbots to recommend them.

When users ask their questions to AI assistants, the systems give “helper” answers with fraudulent links. They visit scam sites by trusting these recommendations and lose their financial information or personal data. Such attacks target users, especially seeking technical support, tax advice or financial services.

Although artificial intelligence companies are developing filtering and verification mechanisms against this threat, new methods that aggressors find quickly evocate. To be careful of users, it is critical to the fact that AI responds are not considered absolute accurate, and especially prefer formal channels for financial decisions.