The EU AI Act is being delayed. Last week, legislators agreed to make significant changes to the AI law, shifting the deadlines for some key requirements. Since the EU’s framework is seen as a global benchmark, its response to industry pressure and competitiveness is likely to shape AI regulation worldwide. Meanwhile, in the U.S., intervention by the Department of Justice (D.O.J) just revamped the Colorado AI bill. Presumably, the delay of the EU AI Act is also likely to add momentum to the federal pushback against AI laws in other states. This is especially interesting because many states that see AI regulation as increasingly urgent are actually pursuing a more business-friendly approach.

The EU is making an effort to balance regulatory oversight with concerns that businesses may not yet be fully prepared to comply with some requirements. Therefore, it is important to understand the decisions EU legislators have politically agreed upon regarding the AI Act. This will help businesses better navigate the regulatory changes ahead.

Breaking Down the May 7 Agreement

The EU AI Act, adopted in 2024, has widely been seen as a global benchmark for AI governance. The comprehensive, risk-based framework has actually served as a model for many other jurisdictions. With that context, the delay in some key requirements from the decided deadline of August 2026 is definitely going to change the conversation around AI governance.

Here are the major changes agreed upon for the act:

• High-risk AI deadlines extended: Requirements for high-risk AI systems, originally due in August 2026, are now pushed to December 2027. For high-risk AI embedded in products covered by EU product safety legislation, the deadline extends further to August 2028.
• Industrial AI carved out: AI embedded in machinery will now be governed by EU machinery rules rather than the AI Act’s high-risk regime. This reduces the compliance burden for industrial operators.
• Safety component definition narrowed: AI systems that are safety components of regulated products are no longer automatically subject to high-risk obligations if their failure does not create direct health or safety risks.
• AI-generated content marking delayed: The obligation to label AI-generated content is pushed to December 2026. This is a four-month extension from the original date.
• New prohibitions on harmful content: Two new categories are added to the banned AI list, which are systems that generate non-consensual sexualized deepfakes and systems that generate child sexual abuse material.
• Expanded use of sensitive data: Processing of sensitive data such as ethnicity and race for bias detection is now permitted beyond just high-risk AI systems, extending to other AI systems as well.
• Broader AI Office oversight: The EU Commission’s AI Office gains expanded supervisory powers over AI systems built on general-purpose AI models and those embedded in very large online platforms and search engines.

The Downstream Effects

The adjusted approach by the EU on AI governance is likely to have changes in how other legislators envision their AI laws. Governments will certainly weigh regulatory oversight against concerns about business readiness and global competitiveness. Here’s what the ripple effects might look like:

Global AI Laws

The EU’s adjustment signals a growing recognition that businesses and public institutions may need more time to prepare for compliance. Therefore, countries that were aligning their own frameworks with the EU might now have less external pressure on several fronts, including high-risk AI systems, rights-based obligations, safety prohibitions, and more. Nations still drafting AI legislation may also shift toward sector-specific approaches rather than horizontal frameworks, taking cues from the refreshed, balanced approach by the EU.

Adding Momentum to Federal Pushback in the U.S.

Many states coming up with their AI laws are already taking a more balanced approach. Still, the EU delay arrives at a moment when the US federal government is actively challenging all state AI laws. We saw how, in Colorado, where the Department of Justice intervened in litigation against the state’s AI Act, there’s now a revamped AI bill passed by the legislature.  Therefore, adjustments in the EU framework give federal critics additional rhetorical ground to frame their AI laws as premature.

California

The interesting thing about California is that when the conversations around the EU adjustments were already getting stronger, Governor Gavin Newsom was tightening his state’s AI standards. I actually talked earlier about Newsom’s executive order challenging the directions by the White House on AI. Rather than broad legislation that can be challenged in court, California is using procurement as its enforcement mechanism. Therefore, for now, it seems California is still an outlier with its own significant influence on the AI governance standards.

How to Move With The Goalposts

The EU’s timeline adjustments buy more time for compliance, but they still don’t change the direction of travel. Here’s what the businesses should keep in mind while navigating the upcoming changes.

• Don’t mistake a delay for a reprieve: Use the extended EU timelines to build compliance infrastructure properly rather than rushing it closer to the new deadlines.
• Map your exposure across jurisdictions: The EU, California, and the federal framework are moving in different directions simultaneously. Understanding where your products and operations sit relative to each other is the first step.
• Take bias seriously regardless of jurisdiction: It’s the one area where regulatory scrutiny is tightening globally, even as other requirements are being adjusted.
• Treat governance as a market signal. Businesses that demonstrate strong AI accountability practices are building trust that translates beyond just compliance and into competitive advantage.

While the Future Takes Its Shape

The EU’s decision to adjust its timelines reflects the complexity of regulating a technology that moves faster than legislation. Notably, the destination still hasn’t changed. Whether it’s Brussels recalibrating, Washington pushing back, or Sacramento holding firm, the signal to businesses is consistent for responsible AI governance. The question is no longer if these rules will affect your business. It’s whether you’ll be ready when they do.