AI Giants Converge on Regulatory Table: OpenAI, Google, Anthropic Dominate AI Rule-Making as Startups Face Existential Crisis

So-called 'regulatory capture' refers to regulatory bodies being overly influenced by the entities they regulate when developing rules, resulting in regulations that reflect large companies' interests more than public interests. In the AI field, this phenomenon is already emerging. OpenAI, Google, and Anthropic not only possess strong lobbying capabilities but also shape regulatory frameworks by providing technical expertise. For example, during the development of the EU's Artificial Intelligence Act, these companies actively participated in multiple consultations, submitting numerous technical recommendations. While these recommendations help improve the technical accuracy of regulations, critics point out they could also be used to set compliance thresholds favorable to large companies.

Specifically, AI regulation's impact on startups is mainly reflected in several aspects. First is compliance costs — large companies can easily afford millions of dollars in compliance expenses, including hiring specialized lawyers, establishing safety evaluation systems, and conducting regular audits, while startups often struggle to afford even these basic costs. Second is time costs — complex certification processes may take months or even years to complete, which is fatal for startups that need to iterate quickly and capture market share. Third is technical barriers — regulators may require models to meet specific safety standards, requiring extensive testing and validation work, while startups typically lack corresponding technical resources and talent. Finally, there's reputation risk — once startups encounter problems during compliance processes, they may face more severe reputation damage than large companies.

The open-source community faces even more severe challenges. Open-source AI models are characterized by transparency, accessibility, and community-driven development, which naturally conflicts with regulation's requirements for centralization, traceability, and controllability. If regulations require all AI models to be certified before release, the release process for open-source models would become extremely complex. Developers might need to apply for new certifications for every model update, which would seriously hinder open-source AI innovation speed. Additionally, open-source models are typically developed collaboratively by globally distributed communities with unclear responsibility attribution, which also creates regulatory difficulties. Some open-source advocates warn that excessive regulation could push AI innovation toward closed ecosystems, harming the entire industry's technological progress.

However, there are also views that regulation is necessary for the healthy development of the AI industry. Without appropriate regulation, AI technology could be used for malicious purposes, or serious safety incidents could occur, which would damage public trust in AI and ultimately affect the entire industry's development. The key is finding a balance — ensuring AI safety while leaving room for innovation. Some experts suggest adopting tiered regulation, implementing more lenient regulatory requirements for small projects and open-source models, while imposing stricter reviews on large commercial models. Additionally, regulatory bodies should establish transparent decision-making processes, ensuring all stakeholders (including startups, open-source communities, and academia) can participate in rule-making, avoiding monopolization of discourse by large companies.