December 17, 2025 | Global Legal Intelligence
Last week showed a worrying trend. The three biggest economies in the world announced AI strategies that are going in very different directions. The US said that federal law will take precedence over state law. The EU pushed down deadlines for compliance while increasing central oversight. China required state data centers to use indigenous chips.
This dispersion makes it hard for Indian tech companies who do business around the world to be compliant. The dream of having the same AI standards all around the world is dead.
## The Fight Over US Federal Pre-emption
President Trump signed an executive order that tried to stop states from regulating AI. The directive wants to make federal law the most important one for AI policy across the country.
The goal of the move is to get rid of California’s comprehensive AI safety laws. It also has an effect on New York and other states that are establishing their own AI frameworks. The government set up an AI Litigation Task Force in the Justice Department. The FCC was told to come up with federal criteria for AI transparency.
But specialists in constitutional law quickly raised red flags. Historically, states have had a lot of power over technology and consumer protection. This is a long-standing constitutional principle.
States with Democratic governors will fight the directive in federal courts. This week, the Attorney General of California will probably talk about the state’s legal strategy. In the first quarter of 2026, several state AGs may work together to file challenges.
The lawsuit might last until 2026 and even into 2027. That means businesses will be stuck in regulatory limbo for years.
### What This Means for IT Companies in India
Indian AI companies with the US as marketplace have to make a very important strategic choice right now. Current state-law compliance frameworks are still legally binding until courts decide what to do about pre-emption problems. That involves following two rules at once.
You have to keep your promises to the state. You also need to get ready for possible federal standards at the same time. This makes governance far more complicated and expensive.
The uncertainty has a big effect on long-term AI outsourcing contracts. It is unclear how to allocate compliance in joint venture structures. Contracts need to include force majeure clauses that take into account big changes in the law.
Indian businesses need to improve their ability to keep an eye on the law. Keep an eye on court cases in different federal circuits. Keep an eye out for motions for summary judgment and preliminary injunctions.
The one good thing is that if federal preemption works, following the rules will be easier in the end. You only have to follow one federal norm instead of 50 separate state ones. But if that happens, that’s at least a two- to three-year time frame.
## The EU’s Strategic Recalibration
The European Commission did things in a very different way. The Digital Omnibus package pushes out the dates for high-risk AI compliance from 2026 to 2027–2028.
This provides businesses greater space to work. The original AI Act deadlines were not achievable for a lot of businesses. The Commission agreed with this fact.
But there is a strategic trade-off that comes with delays. The AI Office gets a lot more power. It now has more power over cross-border AI activities and foundation models. This makes the EU’s central regulator stronger.
The package also makes it clear how the AI Act and the GDPR work together. New rules deal with the problems that come up when you have to minimize data and write a lot of AI documentation. Companies can now better determine whether it is okay to obtain AI training data for valid reasons.
The European Banking Authority gave advice on how to use AI in the financial services industry. Indian Banks with EU operations need to see AI models as more than simply a technical risk; they also need to see them as a conduct and prudential issue. This means that AI used for credit decisions, AML screening, and payments needs to be checked for compliance twice.
### What This Means for Indian AI Development Centers
Indian development centers that work with EU clients have extra time to get ready. But when the deadlines come, they will be watched more closely.
The longer window gives you a strategic chance. Companies who put money into strong EU-calibrated governance now will have an edge over their competitors. People who see delays as a chance to breathe will have problems later.
Another chance is the EU-level regulatory sandboxes. Indian companies can test their compliance methods while being watched by regulators. Being successful in a sandbox shows that you are credible in the market.
But sandboxes need a lot more openness about how AI is made. You have to put up with a lot of regulatory scrutiny. This isn’t for businesses that wish to limit their contact with regulators.
Indian fintechs who work with EU banks are under a lot of pressure. EU bank partners will push compliance requirements down the line. You need to know more than just how to follow the rules for technology. You also need to know how to follow the rules for banks.
The EBA’s approach might change how the RBI thinks about AI governance. If India adopts similar integration principles, banks should expect AI-specific expectations to be built into existing frameworks. Aligning EU regulations ahead of time could make it easier for the US to adapt in the future.
China shuts the door on foreign AI hardware
Last week, China gave a very clear order. Data centers that get money from the government can only utilize AI processors made in the US. Some places were told to get rid of their old NVIDIA and AMD systems.
This is more than just stopping new purchases of foreign chips. China is making it necessary to update old infrastructure. That means that hardware sovereignty is more important than regular technology replacement schedules.
The policy is aimed at data centers that obtain money from the government or do work for the government. Affected facilities must switch to Huawei’s Ascend series and other semiconductor makers in China.
The scope is still not really clear. Does it also cover data centers that only do business? Current reports say that the gradual adoption will last until 2026–2027.
### What Indian Cloud Providers Should Do About This
Indian IT and cloud companies who work with Chinese clients are facing more and more restrictions. You need to figure out if Chinese operations can keep up the level of service with AI chips made in China.
The performance of Chinese chips is different from that of NVIDIA and AMD devices. To do this, you may need to keep two sets of AI infrastructure. Systems for worldwide operations that meet international standards. Localized options for service in the Chinese market.
Indian businesses who use international cloud platforms to deliver AI-powered services to Chinese clients are in a tough spot. If Chinese data residency rules say that processing must happen in China, but the country’s infrastructure doesn’t have competitive AI acceleration, the quality of service goes down.
This means making a big strategic choice. Is it worth spending money on localized technology stacks to stay in the Chinese market?
For a lot of Indian merchants, the economics may no longer support making big commitments to the Chinese market. The decoupling speeds up the consideration of whether the potential for Chinese money is worth the price of building infrastructure at the same time.
There is one possible strategic benefit. The fragmentation opens up chances for AI architectures that work on every platform. Vendors that make AI solutions that work on a variety of hardware platforms get more options. This needs more money for research and development, but it protects against more supply chain problems.
Three Models, No Convergence
Let’s be clear about what we’re seeing. The three biggest AI markets are making rules that don’t work with each other.
The US wants federal preemption, but it will take years of court cases to get it. The EU puts implementation into action with stronger central oversight. China requires that all hardware be made in China.
These methods aren’t just different takes on the same idea. The way they think about rules is very different. They won’t come together.
The US model shows problems with federalism and the First Amendment. The EU approach strikes a balance between new ideas and taking precautions to avoid danger. The Chinese approach puts governmental control and technological independence first.
Every system works toward various political and economic goals. Each one shows a different set of laws and cultures.
What Indian Tech Companies Should Do Now
First, stop hoping that AI regulations will be the same all across the world. It’s not going to happen. Make compliance systems that can handle permanent fragmentation.
Second, put money into compliance designs that are modular and specific to each country. Don’t try to make one method that works for all markets. Make parts that can be put together in different ways for different places.
Third, decide which markets are most important to you. You can’t optimize for all three main markets at the same time. The expenses of compliance and the trade-offs in architecture are too high.
If the EU is your main market, put money into AI Office partnerships and taking part in sandboxes. If the US is important, get ready for years of regulatory instability with flexible contracts. If China is important, think about whether investing in local infrastructure makes sense from an economic point of view.
Fourth, make the regulatory intelligence skills stronger. Each area needs its own monitoring. Court cases in the US. AI Office advice in the EU. Timelines for implementation in China.
Fifth, use the longer EU deadlines wisely. Make compliance skills that set you apart from the competition. As enforcement gets stronger, strong governance becomes a market advantage.
Lastly, think about how this fragmentation may effect your business strategy. Some AI applications can work in places where rules are not clear. Some do not. Platform-agnostic architectures and modular compliance solutions become important tools for businesses.
What to Watch for This Week
United States: The California Attorney General is anticipated to propose a legal strategy for federal AI pre-emption by December 20. Look for a coordinated reaction from multiple states.
European Union: Keep an eye on how member states respond to the AI Office’s increased power. Some states may not want to centralize. Keep an eye on how the European Parliament committees look at Digital Omnibus recommendations.
China: Keep an eye out for more rules about when chip mandates must be put into place. Keep an eye on Chinese AI chip makers’ announcements about how much they can make.
Effect on India: Think about if the changes that happened this week mean that you need to renegotiate compliance milestones in contracts with current clients. Check to see if your current architecture can sustain persistent three-model fragmentation.
The rules for AI around the world just become a lot more complicated. Companies that change their strategy in response will keep their competitive edges. People who want things to eventually come together will have to pay more and lose out in the market.
This Analysis Is About This weekly global legal intelligence report looks at changes around the world that affect cross-border business with India. We keep an eye on changes in regulations in major jurisdictions to help Indian businesses deal with the complicated rules and regulations that apply around the world.
Disclaimer: This analysis gives general facts and strategic points of view. It is not legal advice. Get precise advice on how to follow the law from a certified lawyer.
Next in the series: Tomorrow we look at how financial regulation keeps things in order around the world even though technology governance is falling apart. We’ll also talk about what the changes in sanctions enforcement imply for Indian banks and businesses.
