Elon Musk has called on leading artificial intelligence companies in the US and China to conduct AI safety testing. He proposed that AI companies test each other’s models before releasing them to the public. A cross-checking system could improve AI safety testing without relying on new government regulations.
“xAI, OpenAI, Anthropic, Google, Meta and three or four leading Chinese AI companies should allow rivals to run their safety testing systems on their models,” said Elon Musk, Chief Executive, SpaceX.
Also Read:
Musk Proposes Rival-Led AI Safety Testing
Musk said each AI company already has its own series of tests, or a “test harness”, designed to identify potentially dangerous behaviour.
He proposed that companies use those systems to examine rivals’ models rather than relying only on their own checks.
The idea, he said, would mean companies were effectively checking one another’s work and could raise an alarm before a model reached users.
“Different companies would approach the same model from different angles, increasing the chances of finding problems. The approach can reduce the risk of models becoming too closely adapted to existing benchmarks, a problem that had affected several generations of AI models,” he added.
Musk has previously suggested that leading AI companies should meet every few weeks to discuss safety and security issues, while giving rivals one or two weeks of early access to new models before release. He has said government intervention should remain a last resort if a company fails to address a safety concern raised by another lab.
Also Read:
AI Slowdown Debate Gains Momentum
The proposal comes after leading AI executives have debated whether the industry is moving too quickly.
Anthropic chief executive Dario Amodei has called for a slower pace of AI development, with Musk and OpenAI chief executive Sam Altman backing the broader push.
The debate has also followed warnings from AI researchers.
Former Anthropic employee Jacob Coxon said leading AI labs were “gambling with our lives”, while Anthropic alignment lead Evan Hubinger publicly warned about the potential consequences of advanced AI.
Musk said a cross-company system could be implemented quickly and suggested China could agree to participate. However, he acknowledged that rival AI companies have not agreed to the peer-review proposal.
For the AI industry, the immediate question is whether voluntary AI safety testing between competitors can become a practical safeguard as models grow more capable, or whether governments will eventually be required to set formal rules.
Samarjit Kaur is a journalist and communications professional covering technology & emerging digital trends. With a focus on clarity and context, she reports on developments shaping industries and governance. When not reporting, she chooses to plug-in and relax on her playlists and plan her next bucket-list trips!
