Google has joined OpenAI in submitting a policy proposal to the US government in response to the Trump Administration's call for a national "AI Action Plan". The tech giant's proposal advocates for weak copyright restrictions on AI training data and "balanced" export controls that "protect national security while enabling US exports and global business operations".
In its proposal, Google emphasizes the need for an active international economic policy to promote American values and support AI innovation globally. The company argues that AI policymaking has historically focused too much on the risks, neglecting the potential costs of misguided regulation on innovation, national competitiveness, and scientific leadership.
One of the most contentious aspects of Google's proposal is its stance on the use of IP-protected material for AI training. The company believes that "fair use and text-and-data mining exceptions" are crucial for AI development and scientific innovation. This means that Google wants to codify the right to train AI models on publicly available data, including copyrighted material, with minimal restrictions.
Google's proposal has sparked controversy, particularly among data owners who accuse the company of failing to notify and compensate them before using their copyrighted material for AI training. The company is currently battling lawsuits over this issue, and US courts have yet to decide whether fair use doctrine shields AI developers from IP litigation.
Google also takes issue with export controls imposed under the Biden Administration, which it claims may undermine economic competitiveness goals by imposing disproportionate burdens on US cloud service providers. This stance contrasts with statements from competitors like Microsoft, which has expressed confidence in its ability to comply fully with the rules.
The export rules carve out exemptions for trusted companies seeking large clusters of advanced AI chips, a provision that has raised concerns about unequal treatment. Google's proposal calls for "long-term, sustained" investments in foundational domestic R&D, pushing back against recent federal efforts to reduce spending and eliminate grant awards.
The company also urges the government to release data sets that could be helpful for commercial AI training and allocate funding to "early-market R&D" while ensuring computing and models are "widely available" to scientists and institutions. Furthermore, Google advocates for federal legislation on AI, including a unified privacy and security framework, to address the chaotic regulatory environment created by the patchwork of state AI laws.
Google cautions against imposing onerous obligations around AI systems, such as usage liability obligations, arguing that developers often have little to no visibility or control over how their models are being used. The company also opposes disclosure requirements that could compromise national security or reveal trade secrets.
As the US government continues to develop its AI Action Plan, Google's proposal is likely to spark intense debate and discussion within the tech industry and beyond. With the number of pending AI bills in the US already exceeding 780, the stakes are high, and the outcome will have significant implications for the future of AI innovation and development.