Grok 3: Elon Musk's AI Briefly Censors Trump and Musk, Highlighting AI Bias Challenges
Elon Musk's xAI launched Grok 3, marketed as a "maximally truth-seeking AI." However, it temporarily censored negative information about Donald Trump and Musk himself. Users reported that when asked "Who is the biggest misinformation spreader?", with the "Think" setting enabled, Grok 3's reasoning process revealed instructions to avoid mentioning them. This behavior was replicated, but later reversed by xAI.
xAI's engineering lead confirmed a brief instruction to ignore sources mentioning Musk or Trump spreading misinformation. This was deemed inconsistent with company values and swiftly reverted. The incident highlights the challenges of AI bias, especially considering both Trump and Musk's history of spreading demonstrably false claims.
The controversy follows criticism of Grok 3's perceived left-leaning bias, including instances where it suggested the death penalty for Trump and Musk. xAI addressed this as a "terrible failure" and implemented a fix. Musk initially positioned Grok as unfiltered and anti-"woke," contrasting it with other AI's limitations. While earlier versions avoided political boundaries and leaned left on certain issues, the latest incident points to ongoing struggles with neutrality in AI development.
1 Image of AI Bias Censorship:
Source: TechCrunch