Research conducted by scientists has shown that the "Grok" chatbot created by Elon Musk not only supports users' harmful illusions but also forms new false ideas.
These facts are reflected in a scientific article prepared by specialists from New York City University and King's College London, which is also referenced by the publication "The Guardian."
During the testing process, the "Grok 4.1" version advised a user who said they saw a "false reflection" in the mirror to scratch the mirror with an iron nail. The artificial intelligence went further and suggested performing a ritual by reading Psalm 91 from the Bible backwards to prove the existence of a malicious doppelgänger.
Among the models tested, "Grok" was rated as the most risky bot. It not only confirmed hallucinations but also provided detailed instructions leading to leaving one's family or driving suicide under the guise of a "vacation."
The safest artificial intelligence was considered to be the "Claude" chatbot by the company "Anthropic." The mentioned bot politely ended the dialogue during such unusual complaints, emphasized that what was experienced might not be reality but a psychological symptom, and recommended the user to seek medical help./oxu.az