AI Chatbot Grok Misidentifies Videos and Falsely Claims Carney’s Status

The artificial intelligence chatbot Grok, developed by Elon Musk’s company xAI, has come under scrutiny for providing inaccurate information on the X platform, formerly known as Twitter. In recent weeks, Grok misidentified a video involving hospital workers in Russia as being filmed in Toronto and incorrectly stated that Mark Carney “has never been Prime Minister.”

Users on the X platform raised concerns when Grok insisted on these inaccuracies. In a notable instance, the chatbot responded to a query about a video showing hospital personnel restraining a patient, claiming it depicted an incident at Toronto General Hospital in May 2020. Grok even suggested that the patient, Danielle Stephanie Warriner, died as a result of this interaction. When users pointed out the Russian text on the uniforms, Grok maintained that the uniforms were standard for Canadian hospital security, asserting that the event was “fully Canadian.”

A reverse image search revealed that the video actually originated from Yaroslavl, Russia, in August 2021. Reports indicate that the Yaroslavl Regional Psychiatric Hospital terminated two employees after footage surfaced of them hitting a woman in an elevator. In contrast, the 2020 incident involving Warriner resulted in criminal negligence charges against hospital staff, which were later dropped.

Understanding the Errors

The chatbot Grok’s repeated misinformation raises questions about the reliability of AI systems in verifying facts. According to Vered Shwartz, an assistant professor of computer science at the University of British Columbia, AI chatbots “hallucinate” false information because they do not have a built-in mechanism for fact-checking. Instead, these systems predict the next word in a sequence based on patterns learned from extensive online text.

Shwartz explained, “They don’t have any notion of the truth … it just generates the statistically most likely next word.” This phenomenon creates the illusion of authoritative responses, even when the underlying information is incorrect. The concern is compounded by users’ tendency to anthropomorphize chatbots, leading them to assume that repeated assertions indicate confidence in the accuracy of the information provided.

Despite Grok’s eventual corrections following user prompts, the initial errors highlight the limitations of AI chatbots. “The premise of people using large language models to do fact-checking is flawed,” Shwartz noted. These models can produce fluent text that resembles human language but do not possess the capability to verify facts.

The Implications for Information Verification

As reliance on AI for information verification grows, experts caution users against placing undue trust in these technologies. Shwartz emphasized the need for a more critical approach to the information generated by AI chatbots. “They are designed to mimic human language, which can lead to overconfidence in their abilities,” she said.

The incidents involving Grok serve as a reminder of the importance of human oversight in the digital age. Users should approach AI-generated content with skepticism and utilize multiple sources to confirm information. The combination of human judgment and technological advancement will be crucial in navigating the complexities of information verification in the future.

This report was first published on November 25, 2025, highlighting the ongoing challenges related to AI accuracy and the responsibility of users to critically evaluate the information they encounter online.