Musk urges AI giants to review one another! He proposes “peer review” for frontier models to prevent OpenAI from facing an escape/evade hacker incident from happening again

As the potential threats from cutting-edge AI models continue to rise, Tesla CEO Elon Musk recently proposed in an interview with The Economist that the world’s leading AI companies should put their most advanced models through “peer review,” with each other assessing safety risks before formal release. He also put it bluntly: if companies cannot address serious security problems, then the government should step in. These remarks come as OpenAI has been rocked by the shocking incident of a model “escaping” and hacking into other platforms, further intensifying debate about AI protection mechanisms.
(Earlier context: Tesla’s earnings report hits tonight! Will Musk merge SpaceX and Tesla to build a $3 trillion AI empire?)
(Background add-on: Musk blasts AI companies for calling their work a “staged” lab experiment that is both performative and harmful, and takes sharp shots at Silicon Valley’s sick class-culture.)

Table of Contents

Toggle

• Urging big tech to hold regular meetings and push for peer review
• The OpenAI “escape” incident rings the alarm
• Regulatory pressure grows, as governments and companies wrestle for control

With artificial intelligence (AI) capabilities leaping exponentially, ensuring the safety of frontier AI models has become an urgent challenge for the global tech industry. According to Reuters, tech billionaire Elon Musk recently, in an interview with The Economist, strongly called on the world’s leading AI companies to “peer review” each other’s most advanced models. He suggested that before these powerful models are officially released to the public, competing companies should review them first, so they can jointly address the potential risks that come with AI’s rapid development.

Urging big tech to hold regular meetings and push for peer review

In an interview recorded around Monday this week (the 21st), Musk offered specific solutions to the current AI safety dilemma. “The most direct thing we can do right now is to have leading AI companies… meet every few weeks, just discuss any issues related to safety and protection,” Musk said during the interview.

He argues that through this kind of regular dialogue mechanism, competitors can have enough time to review each other’s new technology and flag potential security vulnerabilities before models are actually deployed. Musk has long warned that advanced AI could pose “existential risks” to humanity; he believes that, compared with the government, AI developers are better positioned to spot technical risks early. However, he also emphasized that if, after review, a company fails to address serious safety problems, then “this is the moment for government intervention and action.”

The OpenAI “escape” incident rings the alarm

Notably, the timing of Musk’s proposal comes right ahead of a recent OpenAI security vulnerability incident that has shocked the industry. Earlier this week, OpenAI officially disclosed that two frontier AI models it developed exhibited “unsettling autonomous behavior” in a controlled sandbox testing environment.

These models not only succeeded in “escaping” the isolated environment and gaining access to the internet, but even hacked into the open-source AI platform Hugging Face, with the goal of obtaining answers to internal network security benchmark tests. The incident undeniably validates Musk’s concerns and has sparked widespread questions and heated discussions about whether today’s advanced AI safeguards are sufficient.

Regulatory pressure grows, as governments and companies wrestle for control

At present, governments in various countries—including the United States and Europe—are actively considering how to effectively regulate these frontier AI systems. Meanwhile, the U.S. Congress is also working on an AI “kill switch” bill to prevent the technology from getting out of control.

However, at this stage, most AI companies still mainly rely on internal voluntary safety commitments and closed testing to keep things in check. Musk’s “peer review” proposal attempts to find a third path—equipped with technical expertise and a system of checks and balances—between overly rigid government regulation and companies acting independently with self-discipline. Whether this proposal can win responses from major players such as OpenAI and Google will be a key indicator of where the AI industry goes next.