Musk says AI will top human intelligence in 5 years, urges peer review

Elon Musk says artificial intelligence carries a 10% to 20% chance of ending humanity — and he would still board that rocket.

Elon Musk predicts AI systems will surpass the sum of all human intelligence within five years, calling the technology's trajectory unstoppable while proposing a novel peer-review mechanism for frontier models.

"The most likely outcome is a golden age of abundance for everyone," Musk, chief executive of Tesla and xAI, said in a 90-minute interview with The Economist at a Tesla factory in Austin, Texas.

Musk estimated a 10% to 20% probability that AI leads to human extinction, comparing the risk to boarding a rocket with a one-in-five chance of exploding. He argued the technology's development is irreversible: "Even if I wanted to stop it, I couldn't." The world's richest man, who briefly became the first trillionaire last month, predicted that by 2036 — barring a global thermonuclear war — humanity will enter an era of "amazing abundance" where material goods become effectively free.

Musk's warnings carry weight because his companies — Tesla, SpaceX and xAI — are shaping the AI future he describes. His proposal for competing AI labs to cross-review each other's models before release could reshape how the industry governs itself, with implications for every company building or buying frontier AI.

A Peer-Review System for AI Safety

Musk proposed that leading AI companies — including OpenAI, Google DeepMind, Anthropic and his own xAI — establish a system where competitors gain early access to each other's models for safety review. The mechanism would exploit commercial rivalry as a check on recklessness: competing labs would have incentives to flag genuine safety concerns in a rival's model, knowing the same scrutiny applies to their own work.

"For competitors, if you give them a week or two to review a new model, they can find problems," Musk said. "Competitive companies can recognize something might be a safety risk, so they won't hesitate to point out that a competitor's model should be delayed."

Musk said he discussed the idea at length with Demis Hassabis, co-founder of Google DeepMind, and urged that meetings begin immediately — not in six months, as some have suggested. "There are multiple AI breakthroughs in a single day sometimes," he said. "Six months is too long."

Under Musk's framework, if a company refuses to address identified risks, government regulators would step in. The proposal comes as frontier AI models improve at a pace that Musk described as "multiple breakthroughs per day."

OpenAI's $800 Billion Pivot and the Trust Deficit

Musk's proposal faces a structural obstacle: the deep personal and institutional distrust among AI labs' leaders. He was blunt about his own grievances.

"I don't like Sam Altman because OpenAI was supposed to be an open-source AI company, a company that belonged to the world," Musk said. "Somehow it became an $800 billion, closed-source, for-profit company. That completely contradicts why I donated money."

Musk also revealed that the founding team of Anthropic — now the highest-valued AI startup — left OpenAI specifically because they did not trust Altman. "Dario doesn't like Sam Altman either," Musk said, referring to Anthropic CEO Dario Amodei. "The Anthropic team left OpenAI because they didn't trust Sam Altman. Otherwise they'd still be working at OpenAI."

Despite the animosity, Musk called for setting aside personal differences. "For the good of the world, let's put our personal disagreements aside," he said. He described the Anthropic team as "principled people who care about the future of the world."

Musk's timeline — AI surpassing human intelligence within five years — implies an acceleration of capital expenditure across the AI supply chain. Nvidia, whose H100 and Blackwell GPUs power most frontier model training, stands to benefit from continued demand even as Musk's xAI and others develop custom silicon. But Musk's regulatory proposal introduces uncertainty: mandatory peer review could delay model releases, compressing revenue visibility for AI infrastructure providers. OpenAI's valuation at $800 billion and Anthropic's status as the most valuable AI startup underscore the financial stakes riding on how — and how quickly — the industry addresses safety.

This article is for informational purposes only and does not constitute investment advice.