Google is asking the federal government to shape AI policy around faster development, broader access to training data, and fewer legal uncertainties for model builders. Its policy proposal aligns with OpenAI on a central issue: whether using publicly available copyrighted material to train AI should be treated as fair use.
The request comes as generative AI systems continue to spread despite high costs and limited profits. The Trump administration has called for a national AI Action Plan, and major AI companies are using that process to argue for rules that protect their ability to build and deploy models.
Google’s copyright argument
Google says it wants “balanced copyright rules,” but the balance it seeks would strongly favor AI developers. The company argues that access to public data, including copyrighted material, is important for improving generative AI systems.
Its concern is not abstract. Google has been accused of using copyrighted data in its models, and it is fighting several lawsuits. The New York Times’ lawsuit against OpenAI could also influence whether AI developers may be held liable for using training data without permission.
Google wants to avoid a future where AI companies must negotiate broadly with rightsholders before training models on publicly available material. The company describes those talks as “unpredictable, imbalanced, and lengthy negotiations.” It also claims that using copyrighted material in AI will not significantly affect rightsholders.
That position puts Google in the same broad camp as OpenAI. Both companies are asking policymakers to reduce copyright risk around AI training, especially when the material is already available to the public.
Data access is only one part of the ask
Google’s proposal goes beyond copyright. The company also wants the federal government to support the physical and institutional systems needed for AI development.
One major focus is energy. Google says AI companies need reliable power to train models and run inference. It projects that global data center power demand will rise by 40 gigawatts from 2024 to 2026, and argues that current US infrastructure and permitting processes are not prepared to meet the needs of the AI industry.
The company also wants the federal government to use AI directly. Google says the government should “lead by example” by adopting AI systems through a multi-vendor approach centered on interoperability.
Its proposal calls for several forms of public support:
- Federal data sets released for commercial AI training.
- Funding for early-stage AI development and research.
- More public-private partnerships.
- Closer cooperation with federally funded research institutions.
- Government-funded competitions and prizes for AI innovation.
Taken together, these requests frame AI as an industry that needs both legal freedom and public investment. Google is not only asking for permission to train on more data; it is asking for the government to help expand the ecosystem around AI.
A national framework instead of state-by-state rules
Google is also pressing for federal legislation that gives AI companies clearer operating rules. The company argues that a “patchwork” of state-level laws could make compliance difficult for developers and deployers.
The source article points to California’s vetoed SB-1047 bill as an example of the kind of state-level effort that worries Google. That bill would have enforced AI safety measures.
Google’s broader message is that national AI policy should prioritize innovation. Its document says, “For too long, AI policymaking has paid disproportionate attention to the risks.” The company wants a framework that supports pushing artificial intelligence further, while assigning responsibilities across developers, deployers, and end users.
That responsibility question matters because generative AI systems are non-deterministic, making their outputs impossible to fully predict. Google opposes efforts to hold model creators broadly liable for how their systems are used. As the company puts it, “In many instances, the original developer of an AI model has little to no visibility or control over how it is being used by a deployer and may not interact with end users.”
Transparency and global regulation
Google is also concerned about rules outside the US. Some countries are pursuing stricter AI regulations that would require more transparency from companies building these systems.
The EU’s AI Act is the example given in the source article. It would require AI firms to publish an overview of training data and possible risks associated with their products. Google argues that such requirements could reveal trade secrets and make it easier for foreign adversaries to duplicate its work.
OpenAI has raised similar concerns in its own policy proposal. Both companies want the US government to push back diplomatically against regulatory approaches they view as too restrictive.
Google wants the ability to release AI products around the world under rules that are lighter and more consistent with what it calls “US values and approaches.” In practice, its proposal asks the federal government to defend an AI policy model that gives developers broad room to train, build, and deploy.
What is at stake
The debate around AI training and fair use is not only about copyright law. It is about who controls the raw material needed to build generative AI systems, who pays for that access, and who carries responsibility when those systems cause harm or create disputes.
Google’s position is clear: it wants publicly available data to remain usable for AI development, a national framework that avoids conflicting state rules, government investment in AI infrastructure, and limited liability for model creators when others deploy or use their systems.
For publishers, artists, researchers, technology companies, and policymakers, the outcome could shape how future AI systems are trained and governed. The national AI Action Plan gives companies like Google and OpenAI a direct opening to argue that copyright enforcement, transparency rules, and liability standards should not slow AI development.