In last week’s LinkedIn article, I wrote about the US White House pivot towards deregulation of AI versus regulatory approaches and highlighted their vision for AI dominance, squarely positioned against China. The USA AI Act was unveiled in July 2025.
I also highlighted that US states are establishing jurisdictional state legislation to ensure that citizens’ privacy and safety are recognized. Bill 896, known as the General AI Accountability Act, regulates California state agencies to assess genAI to critical infrastructure, including potential mass casualty events, with annual reports to the legislature. Although California did not pass its AI Safety Bill (Bill S4017) in November 2024, the debates continue to advance safety controls.
Unfortunately, as these debates flourish globally with inconsistent practices, AI is accelerating even faster as it is being rapidly integrated and embedded into all aspects of our world – business, governments, society, and our everyday lives. We face major strategic alignment challenges and increasing complexity.
As the EU’s AI Act enters into legal status, advancing the world’s first comprehensive legal framework for AI, concerns are mounting on the feasibility of the EU Act to be implementable. However, there is a minimum set of safety standards for AI development, and the big plus is that it eliminates legal uncertainty and can prevent market fragmentation with a unified AI Rule Book to guide behaviors.
Obligations for the EU AI General Purpose AI Models introduced in July 2025 provide a pathway to demonstrate compliance and manage risks. Google has already signed the AI GPAI code of conduct, while Meta has stated it will not sign the code, citing concerns about legal uncertainties and stressing overreach. Other leading companies have signed off on the code, including: Amazon, Anthropic, Apple, Cohere, Google, Microsoft, Service Now, to name a few of the technology leaders rapidly aligned to the new laws.
The code requirements apply from August 2, 2025, and GenAI models must comply by August 2, 2027, giving technology leaders two years to align with the code of conduct. The structure of the code is organized into three main chapters: Transparency, Copyright, and Safety & Regulation, according to the EU Digital Strategy. Each of these files provides helpful guidance to AI companies and AI software developers to approach AI with a stronger commitment to building responsible AI products and services.
Other International AI Highlights
UK
In late July, the UK Government signed a non-binding MOU with OpenAI to promote the public sector adoption of AI, develop safety protocols, and support AI infrastructure. Open AI will collaborate with the UK AI Safety Institute to contribute to emerging AI Growth Zones. The UK government has made clear it is open to US AI investment, having struck similar deals with OpenAI’s rivals Google and Anthropic earlier this year.
India
India is updating its AI sector-specific guidelines, and a new Digital India Act is expected to be released in late 2025. Niti Aayog has reinforced that its guidelines will shape AI for the benefit of the world. India is now shifting from a purely innovation-centric model to one that balances innovation of AI with safeguards, aligning more closely to the EU’s positioning versus the US.
China
China introduced strict new rules in March 2025 mandating explicit and implicit labelling of AI-generated synthetic content. From September 1, 2025, new ‘Labeling Rules’ will come into effect, making it mandatory for AI-generated content to be implicitly labeled, and explicitly labeled where applicable. Explicit labels are those that are easily perceived by users and must be added to text, audio, images, videos, and virtual scenes, while implicit labels are embedded within the file’s metadata.
In July 2025, China has now proposed a global AI governance framework, calling for greater multi-lateral cooperation and warning of the risk of fragmented national strategies.
How to regulate AI’s growing risks was another concern, Li said, adding that bottlenecks included an insufficient supply of AI chips and restrictions on talent exchange. Premier Li Qiang told the annual World Artificial Intelligence Conference in Shanghai on July 26, 2025, “Overall global AI governance is still fragmented. Countries have great differences, particularly in terms of areas such as regulatory concepts, institutional rules,” he said. “We should strengthen coordination to form a global AI governance framework that has broad consensus as soon as possible.”
Conclusion
While we continue to have diverse approaches internationally, balancing innovation and safety. There is hope as conversations emerging from China are warning of the risks of patchwork philosophies, as is India, so the big question is whether the US will demonstrate more international cooperation to balance more safety and control of AI versus being so heavy on innovation first.
We have an imperative for leadership to dig deeper and engage governments, companies, and citizens at large to practice Responsible AI practices.
The impact already to conducting AI business internationally is causing supply chain risks as different governments, procurement offices, and companies are having to be mindful of different jurisdiction positions on AI, and many simply cannot afford to deal with so many levels of uncertainty.
I often say in my speeches, the meat that you put into your sandwiches has more safety controls than your AI models on checks and balances.
Establishing an international AI code of ethics that aligns behavior is a good thing, as it respects humanity and recognizes that we need to be very smart to protect future generations. We have caused so much harm to future generations with inconsistent climate change policies that we now have over 90% of known species extinct.
We cannot stand by – each leader needs to guide their organization forward to manage AI innovation and manage its risks. The risks to our global supply chains are enormous, and as the EU act continues to unfold, and India shortly releases their positioning, we will see if harmonization is feasible in the next 5 years.
It certainly won’t come any sooner.
Here are a few questions to reflect on regarding your organizations.
- Do you have a clearly articulated AI Ethics Policy guiding your selection of AI Use Cases?
- Do you have an AI Governance Center of Excellence (COE) to ensure your organization is developing AI models with prudent transparency, explainability, accountability, and safety practices?
- Do you have digital literacy foundations in place, so everyone can speak the language of AI at a basic level of understanding?
- Have you invested in data management controls to support your AI modelling practices, examples: data lineage, data registration, data labelled, cleansed with verified permission rights?
These are just a few of the questions that need thoughtful practice development.
We have been building AI strategies, AI models, and practices for over ten years and helping many global organizations filter through the complexities and create a sustainable AI roadmap. We have been educating thousands of leaders to advance their AI skills. See link.
The more we learn, we recognize we need to immediately unlearn – this is how fast the world of AI is moving. I have never seen anything like this in my lifetime.
We all have much to do to get AI right, and it starts with each of us making a difference in a way that our voice is heard.
In my mind, your voice matters!
