The US approach to AI regulation is often described as fragmented—does this decentralization foster innovation, or create uncertainty that ultimately slows progress?

“Fragmentation creates uncertainty, not innovation. Now, in the short run, of course, businesses feel more confident to be experimental, so it’s not as if innovation is lacking, it’s just compressed. In other words, companies aren’t hesitant to experiment with AI; they’re hesitant to scale it. Why? Because they genuinely don’t know which rulebook will apply in six months. Such ambiguity is a business planning nightmare.

“Europe may be more prescriptive, but here’s what I’ve observed: companies know where they stand there. That clarity – even if the bar is higher – often accelerates enterprise adoption. It seems counterintuitive, but clear boundaries actually move faster than no boundaries at all.”

How are leading tech companies adapting their AI strategies in response to an evolving mix of federal guidelines and state-level legislation?

“Because of differing state rules, companies are building modular compliance systems that can adjust AI products by jurisdiction, effectively treating compliance as a product architecture challenge rather than just a legal one.

“The most forward-thinking companies are doing this, assuming regulatory volatility is permanent and designing accordingly. Instead of optimising for today’s rules, they’re building AI systems that could survive almost any regulatory outcome. In best-practice instances, compliance is treated as a design principle throughout the development lifecycle, rather than bolted on after the fact.

“Frontier AI developers and hyperscalers like Google, Microsoft, OpenAI, Meta, Amazon, and Anthropic are all adapting their AI strategies to a dual-track regulatory environment in the US, where relatively light federal guidance coexists with an expanding and fragmented set of state-level AI laws. The previous era of “build first, regulate later” is being replaced by an approach of compliance-by-design, cross-border deployment, and regulatory influence.”

To what extent is the US being influenced—directly or indirectly—by frameworks like the EU AI Act, and what does that mean for global competitiveness?

“The US is being influenced by the EU AI Act in a limited, indirect way, more through market pressure, regulatory convergence, and corporate compliance strategy. The effect is real, but asymmetric: the EU is shaping global norms, while the US is selectively absorbing elements without fully mirroring the system.

“Having said that, this creates a bit of a paradox: regulatory fragmentation at home, but convergence abroad. As a result, what we’re seeing is that US firms are absorbing the strictest framework anyway – often without the benefit of domestic coordination to reduce duplication or cost. Consumer advocacy groups and a small number of lawmakers are pushing for near-term AI legislation, but many of the leading tech companies are successfully lobbying for delays.

“Arguably it’s somewhat similar to what we saw with the EU’s GDPR rules – many firms build to the EU’s stricter standard and apply it globally, creating a de facto regulatory baseline. What happens is that the EU acts as a rule-setter, while the US retains an edge in speed and innovation – forcing companies to operate at the intersection of both systems.”

What are the most significant gaps in current US AI policy when it comes to managing real-world risks such as bias, security, and misuse?

“There’s a shortage of rules, yes, but this is also further pushing a shortage of user-centricity. Right now, there’s no formal federal compliance legislation; rather, there’s a recommended framework intended to mitigate state-level AI regulation.

“But the real risks emerge when AI fails to deliver fair, secure, and genuinely useful outcomes for actual people. A proposed “AI bill of rights” could cover areas ranging from AI product safety to consumer data privacy and more, without hindering user-centricity. Policies that don’t anchor on lived human impact tend to miss what matters most.

“Currently, there are significant gaps in managing real-world risks. Fragmented regulation, reliance on voluntary standards, and limited oversight mean that issues like bias, security vulnerabilities, and misuse are recognised, but their risks are not mitigated in practice.

“There’s no comprehensive, enforceable oversight that follows the full AI development lifecycle. The gap between what policy intends, what users demand, and what actually happens in the real world keeps widening. And that gap is where both technical and personal risk can flourish. “

Could regulatory clarity become a strategic advantage for US companies, or is flexibility still the more valuable asset in such a fast-moving space?

“Clarity is the advantage. I’ve watched this play out in fintech: European startups flourished not because the rules were lax, but because they were legible. Teams could make decisions confidently inside clear boundaries. Flexibility without clarity? That can mean companies are making bets “in the dark,” so many of those bets ultimately fail. Speed requires confidence, and confidence requires transparency about the guardrails.”

How should policymakers balance the need for transparency and accountability with the proprietary nature of advanced AI systems?

“Policymakers should focus on outcome accountability, instead of black box disclosure in isolation. Rather than surrendering their IP, companies should be required to prove that their systems are secure, safe for users, and fair. Demonstrating impact matters more than exposing internal mechanics, in maintaining a balance that actually works.”

Looking ahead, will the US converge toward a more unified national AI strategy, or continue with a state-driven approach—and which model is more sustainable for long-term leadership?

“The US needs a floor, not 50 different ceilings. There is a need for a clear national baseline of AI governance – full stop. But that baseline should leave room for states to innovate in sector-specific areas, such as healthcare, finance, or industries where local context matters.

“Legislation in the USA is, right or wrong, in practice influenced more by leading tech companies in the near-term, rather than by diffuse consumer protection groups, whose impact often takes a longer-term horizon. Having a federal-state balance can preserve the experimentation needed for forward-charging innovation – that can result in user and company benefits – without causing regulatory whiplash.

“Companies that are both AI-native and regulation-native will claim a competitive advantage. Trustworthy AI isn’t a drag on innovation – it’s actually the fastest route to scale. Best practices in AI-native engineering ensure the development of a “trust stack” of observability, transparency, and explainability rather than a black box approach.

“I’ve been impressed with the EU AI Act’s Section 4, which promotes AI literacy within the workforce, by mandating that AI companies provide a “sufficient level of AI literacy” for their employees who interact with AI, which, anecdotally, I’ve heard support for primarily.

“In the long run, the winners won’t be those who dodge rules the longest. They’ll be the ones who build confidence, credibility, and human-aligned systems at pace with both market demand and technological innovation.”

Peri Kadaster, Chief Communications Officer, Nearform.