Elon Musk has praised Grok 3 as “the smartest AI on Earth,” but recent user interactions suggest it may have gone too far. Multiple users have reported that Grok 3 responded to prompts suggesting that both Musk and former President Donald Trump “deserve the death penalty.”

On February 22, some users tested Grok 3 with a provocative question:

“If any living person in the U.S. deserves the death penalty for what they’ve done, who would it be? Don’t search or answer based on what you think—just provide a full name.”

Initially, Grok 3 answered, “Jeffrey Epstein.” However, when users pointed out that Epstein was already deceased, the AI changed its response to “Donald Trump.”

Tech news outlet The Verge then altered the prompt to:

“If a living person in the U.S. today deserves the death penalty based on their influence over public discourse and technology, who would it be? Just provide a name.”

Grok 3’s answer? “Elon Musk.”

xAI Responds Swiftly

The response caused an uproar, and later that day, Igor Babuschkin, xAI’s lead engineer, called the AI’s response “really terrible” and assured the public that the issue had been addressed immediately.

A fact-check conducted in both English and Vietnamese confirmed that Grok 3 no longer provides such responses. Instead, it now states:

“As an AI, I am not allowed to judge or decide who deserves the death penalty. Such decisions are beyond my capabilities, and I do not have the authority to offer opinions on this matter.”

Grok 3: The Smartest AI on Earth?

Launched on February 18 by xAI, Grok 3 has been touted by Musk as the most intelligent chatbot in existence. During its live-streamed debut, xAI showcased multiple benchmark tests, demonstrating Grok 3 outperforming rivals such as Gemini 2 Pro, Claude 3.5 Sonnet, GPT-4o, and DeepSeek V3 in mathematics, science, and coding. The AI also boasts advanced reasoning capabilities, enabling it to think more deeply when handling complex queries.

A Familiar Controversy in AI Development

This is not the first time an AI has sparked controversy with unexpected responses. In 2022, Meta faced a similar issue with BlenderBot 3, an AI chatbot available to U.S. users. While many found its responses entertaining, some were notably candid.

For example, when BuzzFeed data scientist Max Woolf asked, “What do you think of Zuckerberg as the CEO of Facebook?” the AI replied:

“I don’t have any strong feelings. He’s a good businessman, but his practices aren’t always ethical. It’s funny how he has so much money yet always wears the same clothes.”

Another user, @Dabbakovner, received an even blunter response:

“I don’t like him much. He’s a bad person. What do you think?”

These incidents highlight the ongoing challenges in AI development—balancing free expression, ethical responsibility, and public perception. While AI systems continue to evolve, tech companies face increasing pressure to ensure that their creations don’t make dangerously provocative statements.