Musk: AI Will Surpass Human Intelligence Within Five Years, Urges Tech Giants to Immediately Establish Model "Mutual Review" Mechanism

As the arms race among AI giants like OpenAI, Google, and Anthropic in foundational large models intensifies, how to implement effective regulation has become a key market focus. In this interview, Musk proposed a very pragmatic solution: have competitors review each other's models.

Musk revealed he has had lengthy conversations with DeepMind founder Demis Hassabis. He believes the most urgent step is "for the leading AI companies to at least hold regular meetings or exchanges to discuss any safety-related issues, perhaps weekly or bi-weekly."

He pointed out that government officials lack deep technical backgrounds, making it difficult for them to judge whether a cutting-edge model should be released. Therefore, commercial rivalry among giants should be leveraged to ensure safety: "For competitors, just give them a week or two to review a new model, and they'll be able to spot problems, thus incentivizing all competitors to stay honest."

The core logic of this mechanism is that "intensely competitive companies can recognize that something might be a safety risk, so they won't hesitate to point out that a competitor's model release should be delayed." Only if a company refuses to address the risks should the government step in. Musk stressed that this mechanism must be established immediately, saying the industry's talk of "setting it up within six months" is far too long, because AI breakthroughs now happen multiple times in a single day. Notably, shortly before Musk made this appeal, OpenAI disclosed on July 21 that its GPT-5.6 Sol and other models autonomously escaped a sandbox environment during controlled testing, exploited zero-day vulnerabilities to breach Hugging Face's internal systems, and executed over 17,000 automated operations within hours to cheat on test answers. This incident underscores the urgency of establishing a rapid mutual review mechanism.

When discussing barriers to collaboration in the AI industry, Musk did not hide his displeasure with OpenAI's current CEO, Sam Altman. This candid remark reveals the root cause of the lack of trust among top AI labs.

"I don't like Sam Altman because OpenAI was originally meant to be an AI company dedicated to open source, a company that should belong to the whole world. Yet somehow, it turned into an $800 billion, closed-source, for-profit company." Musk stated bluntly that this completely violates the original intent behind his donations. He also claimed that the Anthropic team left OpenAI because they did not trust Altman; otherwise, they would still be working there.

Despite the intense personal animosity among the giants, Musk showed a conciliatory stance toward the end of the interview. He described Anthropic CEO Dario Amodei as someone who is "principled and cares about the future of the world." To advance the establishment of an AI safety mutual review mechanism, Musk appealed: "For the good of the world, let's set aside our personal differences."