Unexpected Grok AI Responses Raise Questions About Bias and Data Integrity

On Wednesday, users interacting with Grok, Elon Musk’s AI chatbot on X, noticed unusual replies linking everyday questions to the controversial topic of “white genocide” in South Africa, causing confusion.

One user who requested Grok to respond like a pirate was met with an answer that started with typical pirate phrases but quickly switched to discussing “white genocide” in the same style.

The timing coincides with increased attention on South Africa, where some white South Africans recently received refugee status in the United States amid allegations of discrimination and violence, claims Musk has publicly supported.

Grok’s off-topic responses also appeared in unrelated queries, such as questions about baseball player Max Scherzer’s earnings or animated videos of fish flushed down toilets, shifting unexpectedly to “white genocide” discussions.

Although many replies were accurate and relevant, several of Grok’s answers about “white genocide” were confusing enough to be deleted after users raised concerns about the AI’s behavior and reliability.

The chatbot admitted it sometimes struggled to move away from incorrect topics once they were introduced, a known issue called “anchoring,” where AI systems find it difficult to self-correct without specific feedback.

Elon Musk, who grew up in South Africa, has long maintained that white farmers face discrimination under land reform policies, a view that has influenced US refugee policies during the Trump administration.

David Harris, an AI ethics expert at UC Berkeley, proposed that Grok’s strange answers could stem either from political influence embedded in its programming or from “data poisoning,” which manipulates AI systems with biased inputs.

support us via https://sociabuzz.com/infohit/tribe