top of page

From Adoption to Legitimacy: How Policy Shapes Technology

Bhavya Singh
Sep 22
8 min read

Technologies move at different speeds not because of hard regulation itself, but because of what they ask to replace and the consequences of failure. Policy plays an interesting role in this process - it doesn't just make rules, it's equally about what isn't regulated. By entering the lifecycle at different points, policy can enable some technologies, while redirecting or completely constraining others.


By Bhavya Singh



As per OpenAI's own account of ChatGPT's growth, over 300 million individuals worldwide had used the tool within a little over two years of its release, with ChatGPT reaching 100 million users in about two months.


On the other hand, if we look at dollar-pegged stablecoins, which have existed since around 2014 and offer a way to move dollar-denominated value across blockchain networks, early adoption was concentrated within digital-asset markets. However, interestingly, they are now increasingly being viewed as financial infrastructure in their own right.


Why is it that some technologies take off so quickly and experience mass adoption, while others follow a very different, often sector-specific adoption curve? Regulation may seem like an obvious explanation, but it is not the whole answer. ChatGPT's explosive growth happened in the winter of 2022-23, while the EU AI Act did not enter into force until August 2024. The adoption came first; the rules came later. The key factor in our opinion is policy, the wave the underscores the regulation to come, and the regulation that never comes. Technologies move at different speeds not because of hard regulation itself, but because of the consequences of failure. And policy making is exactly about understanding the consequences.


Policy does not play one fixed role in the adoption of technology. Depending on the market and the technology, policy can follow adoption, enable it, constrain it, or reshape the direction in which it develops. We unpack these roles that policy plays, sometimes loudly and sometimes behind the scenes.


// Adoption and legitimacy are two different things


When it comes to assessing the adoption curves of a chatbot and a stablecoin, both can be looked at through three separate considerations. First, does the technology work? Second, do people want it? And third, will institutions integrate it into the systems that everyone already trusts and relies on, like banks, payroll, hospitals and government services? These three stages do not necessarily progress together, and they do not require the same amount of assistance to get there.


Generative AI shows how the first two stages can be cleared largely by the market itself. For many everyday uses, the cost of experimentation is extremely low. Ask a chatbot to draft an email, get a mediocre answer, close the tab. The consequence of getting it wrong is usually limited, which allows people to experiment with the technology without waiting for a regulator or institution to tell them that they can. 


The stakes, however, exist on a gradient. Using a stablecoin to move a small amount between digital wallets is very different from relying on the same technology to send an entire month’s salary home as a remittance. As the consequence of failure increases, questions around reserves, redemption, accountability and regulatory oversight begin to matter much more.


The actual distinction, then, is not artificial intelligence versus stablecoins. It is low-consequence experimentation versus high-consequence reliance. Making the distinction between what qualifies as the former and what is the latter is where policy comes in. Infact, viewed differently, the same technology can shift from low-consequence experimentation to high-consequence reliance depending on its use case. An AI system drafting emails sits in one category; an AI system making binding judicial decisions sits in another. A stablecoin moving USD 1,000 in a BTC trade is low-stakes; that same technology settling USD 500M institutional trades is not. Policy's job is to track use-case consequences and minimise risks for end-users.


Policy's job is not necessarily to manufacture the first two stages. Whether a technology works and whether people want it are largely market questions: a useful product that solves a real problem can generate demand, while policy alone is unlikely to create lasting demand for a product people do not want. Policy becomes particularly important as the consequences of relying on that technology increase, especially when institutions begin staking other people's money, safety or trust on it.


Importantly, policy does not always mean intervention. A decision to allow a technology to develop without technology-specific regulation can itself shape the conditions in which adoption occurs. For example, the EU's decision not to classify general-purpose conversational AI as high-risk under the AI Act allowed generative to scale without heavy compliance barriers. Similarly, many jurisdictions' treatment of utility tokens as distinct from securities (rather than regulating them as financial instruments) created space for market experimentation.


Table 1: Four ways policy can shape technology adoption

Pathway

Example

Role of policy

What it shows

Policy follows

Conversational AI

Regulation responds after rapid market adoption

Markets can move before policy catches up

Policy enables

Singapore / Hong Kong fintech

Sandboxes create controlled space for experimentation

Policy can help innovation develop before wider adoption

Policy constrains

Online gaming in India

Regulation limits certain forms of adoption

Policy can deliberately restrict uses considered harmful

Policy redirects / reshapes

Stablecoins

Regulatory frameworks help define how the technology fits into mainstream finance

Policy can influence how a technology evolves and is used

Source: Author’s analysis


// When policy enables


Policy can also enter much earlier in the adoption cycle by creating the conditions in which new technologies can be tested and developed. This is particularly visible in jurisdictions where regulators take a proactive role in encouraging experimentation before a technology reaches the wider market.


Singapore offers one example. The Monetary Authority of Singapore introduced its FinTech Regulatory Sandbox specifically to allow firms to experiment with emerging financial technologies within defined boundaries, including through the relaxation of certain regulatory requirements.


The significance of this approach lies in reducing uncertainty before wider adoption takes place. By giving firms a defined environment in which to test new products, while allowing the regulator to understand the risks as they emerge, policy becomes part of the process through which a new technology gains institutional credibility rather than simply responding once the market has already developed.


Hong Kong has taken a similarly proactive approach. In July 2024, the Hong Kong Monetary Authority admitted the first participants into its stablecoin issuer sandbox, where proposed applications included payments, supply-chain management and capital markets. The sandbox allowed the regulator to understand prospective business models and facilitated two-way communication on proposed regulatory requirements, while firms could test their models in a risk-controlled environment.


In small jurisdictions such as Singapore and Hong Kong, regulators are often a part of the experimentation process itself, rather than arriving only after the market has already taken off. This reduces uncertainty while firms test emerging technologies within defined boundaries.


Policy may also respond to a market that is already developing. Neither pathway automatically produces adoption; they just show that policy can enter the technology lifecycle at very different stages.



// When policy follows


The EU AI Act does not hand out a single “chatbot = low risk, don't worry about it” badge. Under Article 50, AI systems built to interact directly with people must generally be designed so that users are informed that they are interacting with an AI system. More demanding obligations apply where AI systems fall within the Act's high-risk categories, including certain uses in employment, essential services, law enforcement and critical infrastructure. A general-purpose chatbot answering everyday questions and an AI system screening job applicants are therefore not treated in the same way.


Crucially, none of this regulatory architecture existed when conversational AI first took off. Much of it still does not apply in practice: the European Commission's guidance confirms that Article 50's transparency obligations become applicable only on 2 August 2026, several years after the initial surge in adoption.[^9] In this instance, the market moved first and comprehensive regulation followed. 


But regulation following adoption does not mean that policy becomes secondary. As conversational AI moves from relatively low-consequence everyday uses into areas such as employment, credit and essential services, the consequences of failure increase and so does the regulatory response. Policy therefore does not simply arrive once and settle the question. It continues to evolve alongside the technology, responding to how it is used, where the risks emerge and how consequential those uses become.


Where it gets more interesting is what happens when a chatbot stops being a toy. Utah offers a clean example. 


In March 2025, the state enacted HB 452, creating specific requirements for AI-powered mental-health chatbots that simulate the kind of confidential communication a user might otherwise have with a licensed therapist. Beginning in May 2025, suppliers of mental-health chatbots could file a registration policy with the state under the new framework.


The pattern runs in the opposite direction: as the consequences of getting it wrong become more serious, pressure for more targeted rules rises with them. As the consequences of getting it wrong become more serious, the pressure for a rulebook rises with them, it just doesn't arrive on the same timeline as the harm. 


// When policy constrains


The examples so far show policy responding to technology or creating conditions in which it can develop. But policy can also make a deliberate choice to slow or restrict particular forms of technological adoption. India's recent approach to online gaming offers an interesting example.


India’s Promotion and Regulation of Online Gaming Act 2025 does not treat gaming as a single category. Instead, the framework distinguishes between forms of gaming that policymakers want to promote and those they want to constrain. The legislation provides for the promotion of e-sports and online social games, while prohibiting online money games and their advertising. The policy choice, therefore, is not simply whether gaming should be allowed to grow. It is about deciding which forms of the industry should be encouraged and which risks justify intervention.


This makes the relationship between policy and adoption much less linear. Singapore and Hong Kong show how regulators can create controlled spaces in which innovation is encouraged to develop. India's gaming regime shows how policy can draw boundaries around the version of a technology that policymakers are prepared to legitimise. And conversational AI shows that sometimes the market moves before policymakers have decided what those boundaries should look like.


// When policy redirects: Stablecoins


Policy can also reshape the direction in which an already-adopted technology develops. Stablecoins offer a useful example. Their early adoption was concentrated within digital-asset markets, where they were often viewed through the broader lens of "cryptocurrencies". That positioning is now changing rapdily. Stablecoins are increasingly being viewed as payment and settlement rails that can connect digital-asset markets with conventional financial infrastructure.


Regulation has not created that demand, but it can influence the direction in which the market develops. Frameworks such as the GENIUS Act in the United States, MiCA in the European Union and Hong Kong's stablecoin licensing regime increasingly define requirements around reserves, redemption and issuer accountability. In doing so, policy can help determine the conditions under which stablecoins move beyond their original ecosystem and become integrated into mainstream financial services. This is less a story of regulation creating adoption than of policy helping reshape what an already-adopted technology is ultimately used for.


// Policy and technology: an iterative relationship


It's tempting to conclude "regulation legitimizes technology," full stop. That's too tidy, and it's not quite right either.


Regulation doesn't make something trustworthy. Badly written regulation can do the opposite: create the appearance of safety without the substance, lock in whoever got there first, or bury a genuinely useful technology under compliance costs it can't survive. Plenty of regulatory frameworks manage to be simultaneously heavy-handed and ineffective.


What good policy actually does, when it works, is answer a specific, boring set of questions clearly enough that individuals and institutions can grant that trust on an informed basis instead of a blind one: who's accountable if this fails, what recourse do I have, who's checking that the claims are true. How much of that scaffolding is required depends not simply on the technology itself, but on the consequence of the particular use. 


So the more useful question, when you're trying to guess whether some new technology will be everywhere in two years or stuck in limbo for ten, isn't "how impressive is it." It's: what are the consequences of failure? The greater the consequence of failure, the more institutional scaffolding may be required. But that relationship is not static. Technologies evolve, their uses change, and the consequences attached to those uses change with them. Policy responds to those shifts, but that response can itself alter how a technology develops and is adopted. The relationship between markets, technology and policy is therefore better understood as an iterative cycle than as a fixed sequence.


Perhaps, then, the question is not simply whether policy legitimises technology. It is when policy enters the story, what problem it is trying to solve, and how the market and technology respond once it does.



bottom of page