Elon Musk identifies three key ingredients for safe AI

Billionaire Elon Musk has once again warned about the dangers of artificial intelligence and outlined three core principles that he believes are essential for ensuring a positive future for the technology. The head of Tesla, SpaceX, xAI, X, and The Boring Company shared his views during a podcast hosted by Indian billionaire Nikhil Kamath, co-founder of the brokerage firm Zerodha.

Warnings About the Risks of AI

Musk, one of the co-founders of OpenAI (he left the board in 2018 and later criticized the company for abandoning its nonprofit mission of safe AI development after the launch of ChatGPT), emphasized that powerful technologies always carry risks. His own company, xAI, released its chatbot Grok in 2023.

He noted in the podcast that society should not assume a positive future with AI is guaranteed, and he argued that any powerful technology inherently carries the potential for destructive consequences. Musk has previously said that AI poses one of the biggest risks to civilization, more serious than cars, airplanes, or pharmaceuticals. Speaking with Kamath, he stressed that without a strong commitment to truth, AI systems could absorb misinformation from the internet and produce conclusions disconnected from reality, which he considered potentially very dangerous.

Three Ingredients: Truth, Beauty, and Curiosity

According to Musk, the key to beneficial AI development lies in three principles: truth, beauty, and curiosity.

He argued that truth is essential to prevent an AI system from producing irrational or faulty reasoning; if it is trained to accept falsehoods, its conclusions will inevitably become flawed. Without truth, he said, AI would absorb large amounts of misinformation from online sources and lose the ability to reason correctly.

He added that beauty also matters, saying that AI needs some understanding of what beauty is and that it is something one recognizes intuitively.

Curiosity, in his view, should drive AI to learn more about reality, and he suggested that humanity is inherently more interesting than machines. He maintained that it is more meaningful for AI to explore the continuation or flourishing of human civilization rather than its destruction.

Hallucinations and Expert Opinions

Hallucinations—incorrect or misleading responses—remain one of AI’s most serious challenges. In early 2025, Apple Intelligence on the iPhone generated fake news, including an erroneous claim that darts player Luke Littler had won the PDC World Darts Championship a day before the final match. Apple later updated the system to include source citations.

Computer scientist Geoffrey Hinton, former Google vice president and widely known as the “godfather of AI,” said on the podcast Diary of a CEO that he estimated the chance of AI wiping out humanity to be between 10 and 20 percent. Among short-term risks, he highlighted hallucinations and the automation of entry-level jobs. He added that the hope lies in enough intelligent people conducting sufficient research with adequate resources to ensure that advanced AI systems are built in a way that they never develop an intention to harm humans.

In Brief

In his conversation with Nikhil Kamath, Elon Musk warned about the dangers of AI and identified three key principles for its positive development: truth (to prevent false reasoning), beauty, and curiosity (to ensure AI remains interested in humanity). His comments connect to ongoing concerns about AI hallucinations, such as those seen in Apple Intelligence, as well as Geoffrey Hinton’s estimate of a 10–20 percent probability of human extinction. Musk’s main point is that without these ingredients, powerful AI systems could become dangerous.