Disclaimer: Originally published here https://techeconomist.hashnode.dev/it-is-more-than-a-blockchain-lessons-from-a-agriculture-traceability-project. Due to personal reasons, this blog has been relocated here.
In 1950, Alan Turing famously asked, “Can machines think?” in his paper “Computing Machinery and Intelligence,” starting the foundational debate on artificial intelligence.
And here we are now, in 2026 with over 19 million AI models, over 8 AI model types, and over 378 million users in 2025. To prevent harms that AI can bring, we need to address advanced cyberattacks (like phishing and malware), AI-powered misinformation and deepfakes, algorithmic bias, and data privacy violations. And on the other hand we need to encourage growth, healthy competition, right for information and innovations. There for more than a decade countries are coming up with AI policies, regulations and act.
For this piece since I am living in the EU, and since US is where almost all the prominent AI companies are developed, I will discuss my perspective on EU AI strategy and US approach to AI.
They both are trying to control. One takes the view of cautious, and careful adoption to AI, where human rights should be valued. The other gives a prominence to innovate and create very powerful AI projects. One values caution, the other innovation.
I’ve been reading : the EU’s AI strategy (2018), and now the simplified version on the EU’s Digital Omnibus(2025) , and the various US regulatory approaches, Biden executive order to foster the AI environment to Trump administration opening up to sell AI hardware to international buyers. I am trying to understand what’s actually happening. And I keep noticing that the good, the bad, the ugly, and most importantly the USEFUL to unearth some complicated underneath.
What are They Actually Trying to Do?
What is important is to understand the logic behind what they are saying.
The EU’s framework is built on a specific threat model. A concentrated power in AI systems can make high stack decisions about people’s lives. Such as manipulation and deception, exploitation of vulnerabilities, and social scoring The EU looked at various scenarios and decided to establish the rules before deployment. This is what they call “ex-ante enforcement.”
The US approach reflects a different bet: AI is new and growing really fast. Predictions are uncertain and the perspective rules can be outdated quickly. Therefore let’s foster theenvironment and hold companies accountable after deployment through litigation, Federal Trade Commission enforcement, liability. This is “ex-post enforcement.”
Here is an analogy. Let’s say you are driving a car, WITHOUT a seatbelt. In the EU approach law enforcement stops you and asked you to wear the seat belt and fines you in case of accident you are unharmed. That is Ex-ante enforcement. In the US approach you never ran in to the law enforcement. You hit a tree and you get injured. The law enforcement will charge you for reckless driving, and you end up in a hospital with a hospital bill. That is ex-post enforcement.
So when people say “Europe is restrictive,” what they actually mean is: “Europe prioritizes preventing harm before it happens.” And when they say “the US is permissive,” they mean: “The US relies on catching and punishing harms after they happen.”
These aren’t just different policies. They’re fundamentally different BETS about what we can prevent versus what we can control through enforcement.
Are they really that different?
Here’s where it gets interesting—and messy.
The US doesn’t actually have “light-touch” regulation, even if it looks that way.
Yes, there’s no comprehensive federal AI law like the EU’s. But the regulatory intensity is there. It is distributed. And it is actively watched by man regulatory guidians. The FTC has become aggressively litigious on AI—investigating algorithmic discrimination, deceptive practices, all of it. And the there is the State of California. The California Privacy Rights Act (CPRA) has granted that effective from July 1, 2023, the users have the right to opt-out of automated profiling and, in some cases, the right to access information about the logic behind AI. On top of all that executive orders on AI safety, procurement standards, algorithmic accountability, the list adds up. The regulatory machinery exists but it’s just operating through different channels.
Meanwhile the EU’s AI Act, while comprehensive, isn’t uniformly strict. It has safe harbours. It has exemptions for research. It has a simpler version for SMEs and SMCs. Has a innovation package for SMEs and Startups. It recognises that a recommender algorithm isn’t the same risk as a hiring system.
The timeline is revealing: I asked Claude to map out when each started thinking about AI policy. The US began about 10 years ago. The EU, 8 years ago. Yet the EU moved to comprehensive regulation faster.
Are there any Real Trade-offs ?
And interestingly, while the EU has been tightening rules, the US has been loosening them. During the Biden administration and also with Trump administration, we see deals between OpenAI, Anthropic along with other companies,(Safety and Testing Agreements (Aug 29, 2024),Department of Defense (DoD) Contracts (June/July 2025), Anthropic Partnership (Nov 2024),The General Services Administration (GSA) agreement (August 2025). This signals a government fostering the AI ecosystem. Further sell AI hardware internationally is not allowed. Meanwhile, the EU raided Grok’s headquarters, considered banning social media platforms, and simultaneously pushed for developing its own competitive AI platforms.
What are the actual trade-offs?
Speed of Innovation: The US approach lets companies move faster. OpenAI, Anthropic, and others have bloomed under this environment. They’re everywhere, and they deliver products quickly. In 2024 Mario Draghi through his report on EU competitiveness pointed how EU has a competitiveness deficit due to its administrative burdens and regulatory inconsistencies. 2024-2029 mandate the Commission pledged to simplify.
Prevention vs punishment: EU invest heavily on prevention. Risk assessments, guidelines, code of practice, that all should follow. It is expensive and requires heavy institutional involvement. US choose punishment. Punishment requires accountability mechanisms that actually work. Involvement of courts, legal systems, and companies solvent enough to pay damages. Both have gaps and comes with tax.
Who pays for all this? EU compliance hits the smaller innovators hardest. So does in US, small innovators will have to defend themselves. Either way small guy has to pay, the giants will have teams and capital to work for them.
I have worked in the SME sector development as a project implementor and my past and current research focus a lot on the BoP, SME and Startups. So the answer for the last question is something I am passionate about.
And to expand more, what about the emerging markets, where 80% of the SMEs in the world are operating? These economies can’t easily adopt the EU model where they lack institutional capacity to enforce complex rules and regulations. They can’t fully adopt the US model either. They lack litigation infrastructure and capital abundance. (Currently, I am doing a whole policy paper on this for an emerging market with some tweaks).
To conclude, in the outset the two approaches look very different, but for some questions they have the same approach. In the whole debate the real questions are taking a back step.
Few articles to get some critique on the different approaches
Europe loosens reins on AI – and US takes them off
Europe Tech Regulation: Yes to Simplification, No to Deregulation
he Global Race to Regulate AI: Biden’s Executive Order Spillover Effects on the EU AI Act
Direct links to acts
Leave a Reply