This month in AI: deployment accelerates, but is regulation keeping up?

Published Oct 31, 2025 · Updated Oct 31, 2025

AI deployment is accelerating, but differing approaches mean regulation is struggling to keep pace

AI deployment is accelerating, but is regulation struggling to keep pace? Image: REUTERS/Dado Ruvic

Cathy Li

Head, Centre for AI Excellence; Member of the Executive Committee, World Economic Forum

New data from the latest State of AI report shows the rollout of generative AI continues apace, with trends emerging around how the technology is being deployed across industries. But is governance keeping up?

In our monthly roundup of AI insights, we look at the latest developments, highlight practical actions from the Forum's new playbook on responsible AI, and track governance and infrastructure moves shaping the ecosystem.

Plus, a chart on how regulatory changes impact AI strategy.

1. Big picture: The world is deploying AI faster than it can govern it - and why that is a systemic risk

The rollout of generative AI continues to accelerate, new data shows. Ramp’s AI Index shows that paid AI adoption among US businesses rose from 5.2% in January 2023 to 43.8% in September 2025, with the technology (73%), finance (58%) and manufacturing (43%) sectors leading the way. The pace of adoption since the start of 2025 has outpaced that of 2023 and 2024.

Advanced agentic and multimodal systems are also gaining traction, with Microsoft, Google and Anthropic among the businesses to have expanded these technologies in the past quarter. The latest State of AI Report, produced by AI investor Nathan Benaich and Air Street Capital, gathered research from a wide range of sources and has explored how these systems are already being deployed. In the healthcare industry, Google’s MedGemma has been shown to improve medical reasoning and understanding of text and images.

But while adoption continues to accelerate, governance of AI is evolving unevenly. Our playbook on Advancing Responsible AI Innovation finds that, despite increased awareness, responsible AI practices by organizations remain immature. Measured on a four-stage maturity scale, a 2025 survey of 1,500 companies found that 81% remain in the first two early stages of responsible AI. This limited maturity is prevalent across organizations, regions and sectors, and shows that although the “why” of responsible AI is largely understood, the "how" remains elusive for most.

Similarly, the Artificial Intelligence Report from Stanford University finds that standardized risk analysis remains rare among developers, and that for companies using the technology a gap persists between recognition and meaningful action. In September, UN Secretary-General António Guterres warned: “Humanity’s fate cannot be left to an algorithm. Humans must always retain authority over life-and-death decisions.”

To complicate matters further, there is wide variance between how different regions are legislating for AI technology. In Europe, phased implementation of the AI Act is underway, and while some obligations are binding, others are in transition. In the US, national and state-level regulations and voluntary commitments create a complex environment. And in China, a sector-specific approach is being taken, with a focus on security and alignment with national priorities.

This week in Kuala Lumpur, the World Economic Forum's Centre for AI Excellence hosted a workshop on building competitive and responsible AI ecosystems across the ASEAN region. One of the key themes throughout the conversations was the importance of strong governance and building trust, as well as the need to invest in the underlying infrastructure that allows AI to scale effectively, beyond just data centres.

With AI technology becoming embedded in critical infrastructure, fragmented regulation and safeguards bring the potential for serious risks.

2. What else is moving in AI

AI-driven shopping is up 4,700% year-on-year, according to the payment services organization Visa. The company is set to roll out a "Trusted Agent Protocol" to help retailers and businesses tell the difference between AI shopping agents and malicious bots. "We believe the entire payments ecosystem has a responsibility to ensure sellers can trust AI agents as much as they trust their best customers and networks," says Chief Product & Strategy Officer Jack Forestell.
Differing approaches to AI infrastructure are emerging globally. In the US, mergers and strategic partnerships - such as OpenAI’s collaboration with Broadcom - are resulting in consolidated efforts to scale the technology at speed. In China, a more distributed approach is being taken with a focus on mixed hardware environments.
Around the world, new legislation to govern AI's implementation have been signed into law. Italy has become the first EU country to pass a comprehensive AI regulation, a bill that aligns with the EU’s AI Act and aims to encourage "human-centric, transparent and safe AI use". Under the legislation, prison terms can be given for those found to have used the technology to cause harm. And in California, Governor Gavin Newsom has signed the Transparency in Frontier Artificial Intelligence Act (TFAIA). The bill includes transparency requirements and reporting of AI-related safety incidents. "California has proven that we can establish regulations to protect our communities while also ensuring that the growing AI industry continues to thrive," he said.

3. Insight in a chart

4. Must reads

Turning measurement into momentum so agile governance can keep pace with AI: This article from the Forum's Kelly Ommundsen, explores which measures can help to build trust in AI by tracking its development amid enforceable systems that provide clarity and confidence
Beyond the desk: How AI is transforming the frontline workforce: In this article, Christof Schminke and Robin Nierynck from Job&Talent uncover how frontline AI could help break the loops of inefficiency and instability, critical in an era of volatile business.
As AI blurs the lines between real and synthetic data, strong governance is essential: DataKind's Lauren Woodman and Arun Sundararajan, Harold Price Professor, New York University, discuss how lines between synthetic and real data are blurring, creating new and significant opportunities, as well as serious risks in this piece.