On 1 August 2024, the European Union’s AI Act entered into force.
The importance of that date was larger than any individual compliance requirement. Europe had chosen to place a horizontal regulatory architecture around a general-purpose technology whose capabilities, use cases and technical standards were still evolving rapidly.
The experiment was not simply whether Europe could regulate AI.
It was whether regulation, standards, innovation and implementation could move at compatible speeds.
Key takeaways
- The AI Act created a risk-based framework rather than one rule for every AI system.
- Entering into force did not mean every obligation applied immediately. The regime was designed around staged application.
- The implementation story became part of the policy story. By 2026, the EU had already modified important timelines through the AI Omnibus.
What happened then
The AI Act entered into force on 1 August 2024 after the EU completed its legislative process.
The framework differentiated between types of AI use and risk. Some practices were prohibited, some systems faced transparency duties, and high-risk systems were assigned more demanding requirements concerning areas such as governance, documentation and conformity.
General-purpose AI also became part of the framework.
The important historical point is that the Act was never a single switch turning every obligation on in August 2024. Its application was deliberately phased.
Interpretation: Europe chose ex ante governance
The European approach reflected a recognisable policy preference.
Instead of waiting for a mature AI market and regulating individual harms afterwards, the EU attempted to establish broad rules for development and deployment in advance.
There is a legitimate argument for doing this. Common rules can increase trust, create a clearer market and establish safeguards before harmful practices become entrenched.
There is an equally legitimate concern: a horizontal rulebook can become difficult to implement when the technology, standards and commercial models evolve faster than the regulatory machinery around them.
The mechanism
Regulation changes business behaviour through cost and certainty.
A firm must identify whether it is a provider, deployer or another relevant actor. It must understand the system’s use and risk category. Depending on that classification, it may need documentation, governance, testing, transparency measures or conformity work.
Those requirements influence procurement, product design, deployment speed and investment.
Good rules can reduce uncertainty.
Unclear or premature rules can increase it.
The strongest countercase
A common criticism is that Europe regulates technologies that other regions commercialise.
That is too simple.
A trusted common framework can itself be economic infrastructure. Clear rules may make adoption easier for customers, public authorities and regulated businesses. A large integrated market can also benefit from a shared rulebook instead of twenty-seven divergent national systems.
The serious question is not “regulation or innovation”.
It is whether the cost, timing and clarity of regulation improve or weaken the conditions for useful innovation.
What changed since then?
In July 2026 the AI Omnibus entered into force, introducing targeted simplification and extending important implementation timelines.
The AI Act reached its general date of application on 2 August 2026, but this still did not mean every provision followed the same date. Certain high-risk requirements were moved later, including to December 2027 and August 2028 depending on the category.
This does not prove that the original Act failed.
It demonstrates that implementation became a moving part of the regime itself.
Scenarios, not forecasts
In a stable implementation scenario, standards and guidance mature and businesses increasingly treat AI compliance as ordinary product governance.
In a continuing simplification scenario, the EU makes further adjustments where implementation costs or missing standards create unnecessary friction.
In a regulatory divergence scenario, major markets develop materially different approaches and global companies must design products around several compliance architectures.
Practical consequences
Businesses operating in Europe need more than a generic “AI policy”.
They need an inventory of important AI uses, clarity about their role in each system, vendor information, a method for identifying higher-risk uses and evidence of who is responsible for deployment.
Employment uses, customer-facing systems and AI embedded in products may raise different questions.
For an international company, this is another reason why jurisdiction cannot be evaluated only through tax. The regulatory environment in which a product is built, sold or deployed affects cost, time to market and governance.
Sources
- European Commission, European Artificial Intelligence Act comes into force, 1 August 2024: https://digital-strategy.ec.europa.eu/en/news/european-artificial-intelligence-act-comes-force
- European Commission, AI Omnibus enters into force, 27 July 2026: https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force
- European Commission, Navigating the AI Act: https://digital-strategy.ec.europa.eu/en/faqs/navigating-ai-act
Disclaimer
This Insight provides general regulatory analysis. It is not legal, technology, employment or compliance advice. The AI Act contains different obligations, roles, exceptions and application dates, and current legislation and guidance should be checked for each system and use case.
