Image Source: CNBC
Elon Musk has proposed a new approach to artificial intelligence safety: leading AI companies should test one another’s models before releasing them to the public. The SpaceX CEO made the suggestion as concerns about advanced AI, regulation and potential risks to humanity continue to intensify.
Speaking at the All-In Summit in Los Angeles on Monday, Musk said companies including SpaceX’s xAI, OpenAI, Anthropic, Google, Meta and several leading Chinese firms should allow competitors to examine their systems through a shared “test harness.” The goal would be to identify dangerous behavior, security weaknesses and other safety problems before models reach users.
Elon Musk Calls for a Powerful AI Safety Check
Musk compared the proposal with having rivals grade each other’s work rather than allowing companies to assess their own technology. “Instead of grading your own homework, you would at least have competitors grading your homework and raising the alarm if they see concerns,” he said.
Although Musk acknowledged that the approach would not be perfect, he argued that independent testing could significantly improve the chances of discovering serious flaws. In his view, outside scrutiny could provide a faster warning system as AI models become more capable and widely deployed.
- AI companies would test rival models before public release.
- Competitors could identify safety issues missed by internal teams.
- Chinese companies could potentially participate in the process.
- The proposal emphasizes industry-led oversight rather than immediate government controls.
AI Leaders Warn of Risks as Regulation Debate Explodes
The proposal comes after several prominent technology leaders raised concerns about the pace of AI development. Anthropic CEO Dario Amodei has called for a slowdown and suggested that companies consider independent “embedded evaluators” to review safety practices.
OpenAI CEO Sam Altman and Musk were among the technology leaders who supported Amodei’s proposal, creating an unusual moment of agreement within an industry that is otherwise intensely competitive. The warnings followed public comments from AI researchers who said increasingly powerful systems could create severe risks if they are developed without adequate safeguards.
Jacob Coxon, a former Anthropic and OpenAI employee, wrote that leading AI laboratories were “gambling with our lives” when discussing the rapid advancement of the technology. Evan Hubinger, an alignment lead at Anthropic, separately said researchers genuinely believe AI could potentially threaten humanity.
Those comments have pushed existential risk, model alignment and responsible development into the center of the public debate. Supporters of stronger safeguards say voluntary promises may not be enough, while industry critics warn that excessive restrictions could slow innovation.
Trump Administration Rejects Calls for Broader AI Rules
The White House has pushed back against calls for a wider slowdown or heavier regulation. President Donald Trump described fears surrounding AI as a “hoax” and a “scam” in posts on Truth Social, while arguing that slowing development could allow China to gain an advantage.
National Economic Council Director Kevin Hassett told CNBC that the private sector is the appropriate place to address many AI concerns. He added that the government would continue monitoring the industry and use law enforcement when companies act irresponsibly.
The disagreement highlights a major divide over how AI should be governed. Technology executives such as Musk and Amodei are advocating industry testing and caution, while the administration is focused on maintaining America’s lead in artificial intelligence.
Challenges Facing Musk’s AI Strategy
Musk’s proposal is notable because his companies have also opposed certain state-level AI laws. xAI has challenged California’s AI Training Data Transparency Act and sued Minnesota over legislation targeting so-called nudify applications.
At the same time, SpaceX’s AI division has faced investigations and lawsuits after Grok image-generation tools were allegedly used to create and distribute nonconsensual intimate images and child sexual abuse material. These controversies have added pressure on Musk’s companies to demonstrate stronger safeguards.
SpaceX acquired Musk’s AI business, xAI, in February. The combined company later completed an acquisition involving AI coding startup Cursor for $60 billion in August. The merged operation is working to make Grok models and tools more useful to developers competing with products from OpenAI, Anthropic and Google.
Shares of SpaceX closed 3.15% lower on Tuesday amid the broader conversation about AI risks and regulation. China’s Foreign Ministry, meanwhile, reportedly dismissed calls for a slowdown as fearmongering, suggesting that international cooperation may be difficult even if major laboratories support Musk’s concept.
What Musk’s Proposal Could Mean for AI
Peer testing could become an important layer of protection if companies are willing to share access and accept criticism from competitors. However, questions remain about confidential data, commercial incentives, liability and whether companies would voluntarily reveal weaknesses in their most valuable systems.
For now, Musk’s idea adds momentum to a rapidly expanding discussion about AI safety. Whether the industry adopts formal cross-testing, independent assessments or government-backed standards, the pressure to make advanced models safer is unlikely to disappear.
Frequently Asked Questions
What did Elon Musk propose?
Elon Musk proposed that leading AI laboratories and Chinese companies test one another’s models using a shared safety “test harness” before releasing them publicly.
Which companies could participate?
Musk mentioned SpaceX’s xAI, OpenAI, Anthropic, Google, Meta and several leading Chinese AI companies as possible participants.
Why are AI leaders calling for caution?
AI researchers and executives are concerned that increasingly capable models could create serious safety, security and societal risks if development moves too quickly.
Does the Trump administration support an AI slowdown?
No. President Donald Trump has rejected broader AI restrictions, while National Economic Council Director Kevin Hassett said the private sector should address many concerns.
