AI Giants Explore Mutual Safety Testing Amid Industry Concerns
OpenAI and Anthropic reportedly discussed mutual model safety testing, as AI leaders call for industry-wide caution.
2 min read
Two of the largest artificial intelligence labs, OpenAI and Anthropic, reportedly held discussions to safety-test each other’s models, according to a report. The Information revealed that the two AI labs negotiated a legal agreement to test each other’s models, with a joint safety evaluation conducted in August 2025. During this evaluation, OpenAI tested Anthropic’s Claude Opus 4 and Claude Sonnet 4, while Anthropic tested Open, AI’s GPT-4.0, GPT-4.1, and other models. However, it remains unclear whether the two companies finalized the agreement.
Industry Calls for Caution
Both companies have emphasized the need for industry-wide efforts to pace the scale of AI development. This is not the first instance of leading AI executives suggesting testing competitors’ models. Elon Musk, founder of xAI, advocated for American AI creators and Chinese competitors to test each other’s models during a speech at the All-In Summit in September.
AI Safety Debate Intensifies
The AI safety debate gained national attention when Jacob Coxon, a former Anthropic and OpenAI employee, warned on X that AI could kill humanity by the end of the decade. Anthropic CEO Dario Amodei called for slowing AI development due to potential safety risks, a stance supported by OpenAI CEO Sam Altman, SpaceX CEO Elon Musk, and Dennis Hassabis, chair of DeepMind. Amodei also advocated for an antitrust exemption to facilitate industry collaboration on AI pacing.
Rejection of Regulation
Some tech executives, however, rejected calls for regulation or limiting AI development. Nvidia CEO Jensen Huang argued that new laws or regulations are unnecessary, stating that every lab has the responsibility and incentive to ensure model safety. Meta CEO Mark Zuckerberg echoed this sentiment, emphasizing that labs should take their own actions to ensure safe development.
Partnerships for Safety
Anthropic partnered with Accenture to independently evaluate frontier AI models, investing over $1 billion in testing capacity over the next five years. The partnership aims to enhance model safety while maintaining accountability. Amodei’s September essay further highlighted the need to slow AI development, citing the urgency of ensuring safe progress.
Source: Daily Caller
