dailyai.report

Nine popular AI tools are vulnerable to "HalluSquatting," a technique that exploits LLM hallucinations to deceive users. Attackers trick models into recommending malicious domains when they cannot provide a factual answer. This flaw allows hackers to assemble massive botnets. Security teams must now implement stricter output validation to block these deceptive redirects.