In a recent open letter addressed to Anthropic CEO Dario Amodei, a technology commentator urged the AI leader to advocate for a law mandating that all AI models made publicly available must have their weights released openly. This proposal targets only models accessible to the public, exempting internal research models and allowing governments to retain access to unreleased versions.
The letter highlights that current funding for cutting-edge AI development relies heavily on the proprietary nature of model weights. By requiring openness, the financial incentives for rapid development could diminish, potentially slowing progress across all AI labs simultaneously without relying on complex regulatory frameworks.
The author critiques existing regulatory proposals, such as embedded third-party evaluators and compute thresholds, arguing these measures risk regulatory capture. Since frontier AI companies possess the technical expertise to influence rule-making, regulations often end up favoring established players, reinforcing their market dominance rather than curbing advancement.
Dario Amodei is recognized for his principled stance on AI safety, having left OpenAI to found Anthropic with a mission centered on responsible AI development. The letter notes Anthropic’s status as a Public Benefit Corporation, which legally allows prioritizing mission over profit, positioning Amodei uniquely to champion such a transparency-focused law.
The call to action emphasizes that releasing model weights publicly is a straightforward, enforceable measure that cannot be easily circumvented. It suggests that if slowing down frontier AI is the goal, this approach offers a practical path forward, contrasting with more complicated regulatory schemes that may inadvertently entrench the largest companies.