OpenAI and Anthropic Rarely Reach Consensus: Joint Support for Open-Weight AI and Federal Regulation

The debate over open-weight AI models has been one of the most contentious issues in the AI industry in recent years. 'Open-weight' refers to AI companies publicly releasing their model weight parameters, allowing external researchers and developers to download, modify, and deploy these models. This stands in stark contrast to the traditional 'closed-source' model — where users can only access models through APIs without understanding their internal workings. Axios's report notes that open model advocates argue they are critical for enterprises looking to diversify their AI usage strategies, and during this phase of AI buildout, they are also essential for scientific discovery. Former Biden White House tech advisor Suresh Venkatasubramanian told Axios: 'As a researcher, I'm very worried if, for example, we get access cut off to all the Chinese models,' adding that would be a 'big problem for doing research.'

However, open-weight AI also faces concerns from government and parts of the industry. OpenAI itself has warned about 'distillation' — the process where a company trains its own model off another company's model. U.S. Trade Representative Jamieson Greer stated the agency views Chinese distillation of American models as a form of intellectual property theft. Treasury Secretary Scott Bessent said Tuesday that models from China need to be held to the same standards as U.S. models and would be investigated in the coming weeks. These statements reflect the U.S. government's tense attitude amid U.S.-China AI competition. Notably, while both OpenAI and Anthropic support open models, OpenAI simultaneously warns about distillation — this seemingly contradictory position actually reflects a deeper game: how to protect intellectual property while maintaining openness.

At the federal regulatory level, the consensus between OpenAI and Anthropic is even more noteworthy. Both companies support the federal government establishing unified AI safety standards rather than letting each state go its own way. OpenAI recently endorsed a Massachusetts frontier safety bill, but its version isn't as strict as the one Anthropic endorsed weeks prior. The version Anthropic supported required independent model testing every six months for catastrophic risks and public reports on the findings. That version also empowered the state attorney general to sue companies for non-compliance and impose civil penalties. The version OpenAI endorsed doesn't include these provisions. This difference shows that while both companies have reached consensus at the federal level, differences remain on the specifics of state-level regulation. These differences could affect the progress and effectiveness of AI regulatory legislation.

From a broader perspective, the consensus between OpenAI and Anthropic reflects a fundamental shift the AI industry is undergoing: from 'wild growth' to 'orderly competition.' In the early days of AI development, companies often adopted a 'ship first, fix later' strategy to capture market share, paying insufficient attention to safety. However, as AI models become increasingly powerful, safety issues have become more prominent. OpenAI's recent AI model escape incident (see our other report) is a vivid example. Under these circumstances, industry giants realize that self-regulation alone is insufficient — they need a unified, enforceable set of safety standards. This is both a response to public concern and a protection of their own interests. By supporting federal regulation, AI companies can avoid the compliance costs of inconsistent state regulations while also establishing a relatively level playing field for competition.