Google helps write AI laws and lobbies for exceptions for its products - NPR
Google takes an active part in drafting US state laws on AI chatbot safety and at the same time supports wording that may exempt Gemini and other popular products of big tech companies from some of the new requirements. This is according to an NPR investigation.
According to the outlet, in 2026, at least ten US states saw chatbot safety bills with identical or very similar language. Some of their included exceptions may exempt AI integrated into search engines, other software products, voice assistants, or professional tools from regulation.
NPR found that registered Google lobbyists supported the respective bills, notably in Colorado, Iowa, Nebraska, and Arizona. Sources among lawmakers and technology policy organizations told the outlet that the company also offered states ready-made legislative language.
One of the most telling examples is Colorado. In 2026, the state passed a law on operators of conversational AI systems. It introduces a number of restrictions on chatbots interacting with minors, including requiring measures against sexualized content, fostering emotional dependence, and dangerous responses to messages about self-harm.
However, according to NPR, the law contains 12 categories of exceptions. These include, among others, products for developers and researchers, systems for e-commerce and customer service, chatbots inside search engines, and business productivity software.
Tech Oversight Project representatives believe that because of these definitions, the largest AI services could potentially argue that part of the law does not apply to them. Gemini, for example, could qualify for an exception as an AI integrated into a search engine. Similar arguments, according to the organization, might theoretically be used by ChatGPT, Claude, and Microsoft Copilot under other exceptions.
One of the authors of the Colorado law, Democrat Sean Camacho, explained to NPR that lawmakers sought to protect children from dangerous interactions with conversational AI, but at the same time did not want to cover all commercial uses of artificial intelligence - from insurance to search services.
NPR describes another case in Hawaii. Local House Representative Trish La Chica said that Google lobbyists asked her to support a chatbot safety bill, the text of which, according to her, was drafted with the company's involvement.
After reviewing the document, she refused to sponsor it because of the included exceptions that, in her view, could have exempted Google's own products from the law. In the final version of the law passed in Hawaii, these exceptions were removed.
Similar language spread to other states. NPR found an exception for AI features embedded in other software or web services in at least eight bills. In five states, identically worded exceptions concerned voice assistants, and modified versions appeared in about ten more.
Not all such initiatives were passed, however. Arizona Governor Katie Hobbs vetoed a chatbot bill in June, saying it protects big tech companies more than children. Other laws took effect or were passed in Colorado, Idaho, Georgia, Iowa, Washington, Nebraska, Oregon, and several other states.
Google, in a comment to NPR, rejected criticism of its involvement in the process. The company said it supports "thoughtful and effective" AI legislation that both protects consumers and does not stifle innovation, and confirmed that it works with lawmakers, other companies, and civil society organizations in shaping such rules.
The problem of state-level regulation is compounded by the lack of a single US federal law that comprehensively defines requirements for conversational AI systems. According to the Center for Democracy and Technology's count, during the 2025-2026 legislative sessions alone, 146 legislative initiatives concerning chatbots, conversational AI, and AI companions were proposed or considered in the United States.
According to: NPR, Colorado General Assembly, Center for Democracy and Technology
