Google’s Kent Walker argues IP rules must adapt for AI while protecting creators

Alphabet and Google’s top global affairs executive used a keynote in Singapore to argue that intellectual property rules need to keep pace with AI, while still protecting creators and encouraging innovation. Kent Walker, President of Global Affairs for Google and Alphabet, delivered remarks at Singapore IP Week 2026 framed around “the breakthrough in how we make breakthroughs,” describing generative and agentic AI as a shift in the way new discoveries are produced and deployed across industries.

Walker said Google’s mission is to organize the world’s information and make it universally accessible and useful, and he cast AI as a “quantum leap” toward that goal. He pointed to improvements he said came quickly in recent years, asserting that Google’s new AI models are “300 times more efficient than those from just two years ago,” and argued that the larger impact is how AI is being used to deliver results in real environments rather than simply making predictions.

The keynote also emphasized what Walker described as a change from today’s systems acting only as tools that answer prompts. He said newer models can “work independently,” take different actions in different environments, and “course correct when they run into dead ends.” In his view, this capability is contributing to scientific breakthroughs in areas such as medicine, energy, and materials science, including what he called breakthroughs in “how we make breakthroughs.”

Walker offered examples of AI accelerating research and practical outcomes. He cited University of Oxford researchers using AlphaFold, described as a Google-built system that predicts the 3D structure of proteins, to design improved malaria therapies, including a new vaccine he said is more effective. He also described AI as an “everyday” force multiplier for small businesses and entrepreneurs, arguing that adoption in Asia-Pacific is supporting this broader diffusion of AI into daily economic activity.

To illustrate the downstream business impact, Walker referenced a Japanese travel platform called ikura that he said uses Google AI to help tourists find experiences beyond major tourist areas, with the goal of supporting small businesses that might not otherwise reach travelers. He also pointed to Aruna, a startup in Indonesia that he said uses AI to help fishermen in remote villages get fairer prices for their catch, describing the effort as reducing barriers such as logistics and language constraints.

Walker argued that enabling environments matter because the competitive edge increasingly comes from deployment and integration, not solely from inventing new technology first. He said Asia-Pacific already shows high levels of AI adoption, describing India and Indonesia as having the most absolute users and Singapore as having some of the highest adoption per person. He added that he said 29% of AI users in Singapore are “super users,” a category he defined as people able to delegate whole tasks, automate daily routines, and use paid models. He further stated that around two-thirds of people surveyed in the region are optimistic about AI, and he tied that optimism to what he described as the strategic advantage of putting AI to work.

The core of Walker’s keynote focused on how AI and intellectual property systems should interact. He said capturing AI-driven gains depends on countries creating the right enabling environment, and he described Google as commissioning an Oxford Economics report that examines how copyright rules influence how much value countries capture from AI. He then posed the balancing question he said the conference must ask: “Can we safeguard creator rights while still delivering on the promise of AI?”

On patents, Walker argued that the system is already supporting AI development and deployment, and he said Google has the “largest AI patent portfolio” with foundational patents. He acknowledged a surge in AI patent applications, arguing that the problem is compounded because generative AI makes it easier to draft applications. He cited a World Intellectual Property Organization observation that more new patents were published in 2024 and 2025 than in the prior ten years combined, and he said claims are appearing across AI model architectures and novel applications. Walker said patent offices may need tools to evaluate claim quality, including potentially using AI to analyze prior art and to “deconstruct claims generated by AI,” while maintaining the overall architecture where it works and adapting to rapid change.

On copyright, Walker said the conversation is more complex because traditional copyright has focused on outputs, not inputs. He said Google is prioritizing safeguards for AI outputs, including deploying advanced, modality-specific filters to help prevent models from exactly replicating content that might be present in training data and using notice-and-removal systems to remove allegedly infringing material. He argued that existing copyright principles are robust, insisting that if a user creates an infringing output, it is infringing regardless of the technology used, and he said liability should draw a line between the tool and how someone chooses to use it.

Walker also addressed how AI models are trained. He said training can be analogized to how human creativity draws on what came before. As an illustrative analogy, he compared generative AI training to students reading books in a public library, learning craft from prior works, and then writing original novels without infringing. He argued that generative AI training operates in a similar pattern-recognition way. He warned that a regime requiring commercial licenses for every piece of publicly available data used for training would end AI innovation, and he advocated for text and data mining exceptions for training on publicly accessible content, pointing to approaches he said were pioneered by Singapore, Japan, and the EU, and noting that he said courts in India have followed a similar approach.

Looking ahead, Walker said balanced rules also recognize content’s scale on the internet, giving publishers and creators the ability to opt out of having their content used to train or ground model outputs. The keynote left unanswered details about exactly how any specific patent or copyright reforms should be implemented across jurisdictions, including how policymakers should treat edge cases where training data, output behavior, and attribution rules intersect. Still, the message was clear: AI’s acceleration is raising IP pressure points, and countries that can align legal standards with AI development and deployment may be better positioned to capture the economic value Walker described.

In the near term, observers will likely watch how governments and IP stakeholders respond to Walker’s calls for updated training rules and practical mechanisms for assessing patent claim quality in a world where AI can generate filings. The conference setting also suggests renewed attention to notice-and-removal processes, opt-out frameworks, and the extent to which AI systems should be treated as tools versus actors when allocating rights and responsibilities. Whether lawmakers converge on a common baseline for text and data mining exceptions, and how effectively they can police output copying, may determine how quickly AI innovation can scale without reigniting creators’ concerns.