In President Trump’s thousand-page-long “One Big Beautiful Bill,” a last-minute curveball was slipped in by the House Energy and Commerce Committee: a 10-year freeze on any state-level regulation of artificial intelligence.

“No State or political subdivision thereof may enforce any law or regulation regulating artificial intelligence models, artificial intelligence systems, or automated decision systems during the 10-year period beginning on the date of the enactment of this Act,” the bill states. 

At the time of this writing, the bill had passed the House, and was being reviewed in the Senate.

A decade is an eternity in AI land. Imagine a world where, free from oversight but guided by the natural checks of market forces, AI develops innovative, cost-effective and scalable carbon sequestration technologies that reverse the damage of climate change and restore balance to our planet. Equally, imagine if unregulated AI lands us deep into a post-truth world, where we have learned to blindly trust it because figuring out fake from real is just too hard, and where we allow silicon- based intelligence to even shape our beliefs. Or imagine a future in this unregulated AI world where, somehow, we have all the benefits without any of the downside.

You might have come across this riddle: “Why do cars have brakes? So you can drive fast.”
 Speed without control is a like a car without brakes, a toddler with a chainsaw — or beta-testing a new vaccine on the entire population.  

In the world of AI, we’ve got all the speed we need. But there are no brakes yet, and this bill proposes none for 10 years. Can market forces help the system self-regulate?

Perhaps — at least to a certain extent. Sitting at the interface of business and consumers are well-established regulatory protections. As regulated entities, enterprises can, and sometimes legally must, shield consumers from any harm caused by the unregulated AI they use in their products and services. There’s some comfort in that.    
 The frame of commerce and economics is too small to accommodate the full impact of brakeless AI. We must extend our system to include society, politics and the environment. Even if all existing consumer protections remain in place, unchecked AI will create joblessness at a scale that we can neither compute nor fathom. Whose problem is that?  We can’t leave it for market forces to resolve.

A Regulatory Timeout?

The Tax Cuts and Jobs Act aims to ignite American innovation, with AI at its core. But while the bill promises R&D credits, depreciation perks and infrastructure boosts, it also prohibits state governments from imposing any guardrails for 10 years. The only brakes we’re left with are voluntary self-regulation by AI makers who are in the middle of an escalating arms race, and (2) the invisible hand of market forces.

The intent, supposedly, is to avoid a patchwork of conflicting state laws. But in doing so, as Fierce Healthcare points out, the bill even overrides state protections already in place to guard against algorithmic bias, opaque decision-making and consumer harm.  A group of 13 state attorneys general from both parties has formally objected, warning that it threatens civil rights and consumer protections.

Looking back on the evolution of social media, many would agree that, had we regulated around privacy, transparency and mental health impacts early, we’d be in a different place. Instead, we got chaos, polarization — and ad revenue. Even Meta’s Mark Zuckerberg, once a regulation skeptic, now calls for “smart guardrails.”

Now we are in the infancy of generative AI, with far deeper and wider implications than social media. Its problem-solving capabilities went from kindergarten to PhD levels in the blink of an eye. In a decade, we’ll be living in systems shaped by AI. Decisions about insurance eligibility, medical prioritization, even disaster-relief logistics will increasingly be influenced, if not determined, by AI.

We have no idea where AI is headed next, not even the companies developing it. Is this the best time to apply a “figure it out as we go” or “wait and see” national strategy?   AI is already making decisions with major human consequences.  

Removing Oversight Won’t Solve It 

The bill is generous to AI companies, with the goal of supercharging the technology. But it’s unlikely to be achieved with speed alone. The bottleneck for enterprise adoption is trust, transparency and accountability. We’ve seen this even for technologies that are proven to work far more reliably than humans, like the autonomous vehicle. We need to build a high level of trust before adoption at scale, especially for cases where the cost of a mistake is high.  

In another universe, this bill might not have caused much concern. If the dominant AI and language model architecture didn't guess or hallucinate, but instead was built from the ground up on a verifiable model, we might have been OK with a lower level of regulation.

The state of GenAI today is quite the opposite. For all its brilliance, it’s fundamentally unexplainable. Decisions aren’t verifiable and outputs may be fabricated. And if the output of a “black box” leads to serious consequences, how do you troubleshoot? Whom do you hold accountable?
 Transparency is foundational — not just for debugging or data science hygiene, but for legal, ethical and operational accountability.

In AI land, the incentives are strong, the market hot, and the instinct is to race ahead. But the bottleneck right now isn’t speed. Moore’s Law is accelerating everything, and AI makers are dragging all these issues with them into future models. 

Transparency and explainability aren’t luxuries — they must be infrastructure. It’s how we earn trust, ensure accountability and raise levels of AI adoption in the enterprise.

Making Smart AI Policy

We don’t need to choose between chaos and paralysis. There’s a middle path.

The American Action Forum supports federal-level guidelines that are clear, adaptive and unified, without freezing out state innovation entirely. The Wilson Center reminds us that this isn’t just about industry; it’s also about national security, international leadership and long-term democratic resilience. The Center for Strategic and International Studies’ AI diffusion model shows how tiered global access to frontier AI can preserve strategic leadership while avoiding techno-isolation, or an all-or-nothing regulatory stance.

The EU AI Act focuses on use-case risk, not the technology itself. That’s the kind of nuance we need — something between a 10-year hall pass and stifling regulation.

The systems we build today will shape everything from public health to disaster response. We can’t afford to get this wrong. We need to play the long game, where AI is fast and reliable, adopted and governed.

Joy Dasgupta is chief executive officer of Gyan.