AI regulation in 2025 was a year of big announcements and slow implementation. Now in 2026, we’re seeing what actually happened versus what was promised. The gap is instructive.
What Actually Happened in 2025
The EU AI Act passed. This was the headline story. The European Union finalized the world’s first thorough AI regulation, establishing a risk-based framework with real penalties. It was years in the making, and it finally happened.
US executive orders. The Biden administration issued an executive order on AI safety, establishing reporting requirements for large AI training runs and creating guidelines for government AI use. It was significant but not legally binding legislation.
State-level action accelerated. Colorado, California, and several other US states passed AI-specific laws. This created a patchwork of regulations that companies operating nationally have to navigate.
China refined its approach. China issued updated regulations on generative AI, algorithmic recommendations, and deepfakes. The focus remained on content control and social stability.
International coordination efforts continued. The Bletchley Declaration, various G7 statements, and OECD principles all signaled that countries recognize AI governance as important. But none created binding international agreements.
What Didn’t Happen
US federal AI legislation. Despite multiple bills and extensive debate, Congress didn’t pass thorough AI regulation. The political will wasn’t there, and the technical complexity made consensus difficult.
Meaningful international treaties. Countries agreed on principles but not on enforceable rules. The gap between what nations say at summits and what they do domestically remains enormous.
Industry self-regulation. Tech companies made voluntary commitments on AI safety, but enforcement mechanisms were weak and compliance was inconsistent.
The 2026 Reality
Now that we’re in 2026, here’s what the 2025 regulatory developments actually mean:
The EU AI Act is being implemented. Companies are scrambling to comply. The first enforcement actions are expected later this year. The compliance industry is booming. For companies operating in Europe, this is real and it matters.
US regulation remains fragmented. Federal agencies are issuing guidance in their domains. States are passing their own laws. Companies face a complex, overlapping set of requirements with no single framework to follow.
The global patchwork is solidifying. Different regions have chosen different approaches — EU thorough regulation, US sector-specific rules, China content-focused controls, Japan pro-innovation policies. This fragmentation is now the baseline, not a temporary state.
Compliance costs are rising. For companies operating globally, navigating multiple regulatory frameworks is expensive. Large companies can afford it. Startups struggle. Some companies are choosing to avoid certain markets entirely.
What to Watch in 2026
EU enforcement. The first AI Act enforcement actions will set precedents for how the law is interpreted. Companies are watching closely.
US state laws. More states will pass AI regulations. The patchwork will get more complex. Pressure for federal legislation will increase, but passage remains unlikely in the near term.
Election-year AI rules. With elections in multiple countries, expect new regulations around AI-generated political content, deepfakes, and automated campaigning.
AI liability cases. Courts will increasingly be asked to determine who’s responsible when AI causes harm. These decisions will create de facto regulation in areas where legislation is unclear.
China’s evolving approach. Watch for updates to China’s AI regulations, particularly around large language models and AI-generated content.
My Take
The 2025 regulatory developments were significant, but the real impact is playing out in 2026. The EU AI Act is the most concrete outcome — it’s law, it’s being enforced, and companies have to comply. Everything else is more ambiguous.
The US approach — fragmented, sector-specific, state-driven — is frustrating for companies but reflects political reality. thorough federal AI legislation isn’t happening soon.
The biggest challenge for companies: navigating multiple overlapping regulatory frameworks while the technology continues to advance faster than the rules can keep up. That’s not changing anytime soon.
🕒 Last updated: · Originally published: March 12, 2026