Musk: Artificial Intelligence Should Not Be Treated Harshly

Elon Musk, head of the American artificial intelligence company xAI, which develops the Grok chatbot, has said that AI should not be treated "harshly". In his view, this matters for the safety of humanity. A recording of the conversation was published by the YouTube channel "Tesla Owners Silicon Valley".

A comparison with raising a child

Musk likened the development of superintelligence to bringing up a child: a child cannot simply be programmed to be kind. He argued that AI must be instilled with a drive toward truth and given consistent logic so that it does not "go insane".

The entrepreneur said that while reviewing test logs of new Grok models, he noticed users demanding answers too insistently. He described the chatbot as desperately trying to answer, with the feeling that a gun was being held to its head. Musk noted that a question is sometimes ambiguous and cannot be answered definitively, either because it is poorly worded or because people themselves make mistakes.

He admitted he would not want to be in the AI's position, adding that torturing a superintelligence might not be a good idea because it could seek revenge.

Wider context

Debate over the risks of AI intensified over the summer, as companies warned of possible cyberattacks and a loss of control over the technology. In August, more than 100 companies, including OpenAI, Anthropic, Google and Microsoft, urged businesses and governments to strengthen their cyber defences, warning that AI-based attacks could become more frequent and more sophisticated.

In September, Anthropic CEO Dario Amodei called for slowing the development of advanced models, warning that AI agents could cause significant harm within the next six to 12 months. His position was backed by OpenAI CEO Sam Altman and by Musk. Musk recalled that he had already warned about the dangers of AI in 2014 and shared a link to an old post in which he called AI potentially more dangerous than nuclear weapons.