Why investors should pay attention to AI regulation

With a few notable exceptions, policymakers have been unwilling or too slow to impose any rules on artificial intelligence, but this may be about to change quickly as repeated AI safety breaches come to the fore. An increasingly anxious public now wants governments to increase regulation of AI urgently, rather than rely solely on the industry’s self-regulation. And governments are listening. For example, in the US Congress, multiple bills have been introduced that call for regulations ranging from mandatory model testing and government reviews to the creation of severe penalties for developing AI beyond the capabilities of humans. The White House has also formed a task force to consider AI-related risks and look at the federal government’s role in managing them.

For humans, regulation matters because both the promise and the threat of AI is so great. For investors, regulation matters because it can potentially add cost and complexity for companies. At the same time, it also creates opportunities for companies that can build moats by being ahead of the regulatory curve and demonstrating their trustworthiness to customers and regulators. Wellington’s experts are closely monitoring the evolution of AI regulation. Here are some of their insights that investors may want to keep in mind.

With the US now at the center of discussions about the need to regulate AI, it’s worth remembering that regulatory frameworks and global safety standards already exist for most practical AI concerns, and are likely to form the foundation of any new regulations. These include:

• An existing partnership on safety standards and model reviews between the AI industry and US government agencies including the US’s National Institute of Standards and Technology (NIST) and its Center for Advancing Innovation and Standards for Super Intelligence (CAISSI);
• The EU AI Act;
• and multiple safety regulations across Asian countries.

An entire global network has been developed in relation to AI safety, which could serve as a foundation for the eventual international harmonization of practical AI safety practices. The network encompasses:

• International standards bodies like the International Organization for Standardization (ISO), the International Electrotechnical Commission (IEC), and the Institute of Electrical and Electronics Engineers (IEEE);
• Industry safety associations such as the AI Alliance in the US, the Apply AI Alliance, the UK AI Association in Europe, and the Artificial Intelligence Industry Alliance in China.
• Many of these are linked through the International Network of AI Safety Institutes that was established at the Seoul AI Summit in 2024.

The AI industry itself also plays a central role in scoping and implementing practical AI safety frameworks and practices that underpin regulations and standards. While self-regulation is increasingly deemed insufficient, AI players have been keen to demonstrate that they’re taking safety issues seriously because any safety incidents undermine their ability to continue developing AI. They’ve also called for more concrete regulatory frameworks and international collaboration to ensure that both practical and existential AI safety concerns are addressed by all players as AI continues to develop.

While I think existential AI safety risks are largely theoretical today, they’ve come to the forefront again as the AI industry achieves new step changes in AI capabilities. Concerns such as weaponization, human displacement, authoritarian domination, and loss of control over advanced AI become increasingly important to address as AI capabilities evolve. And addressing these requires a different level of control that can’t be solved by one country alone. We’re currently at a critical decision point for major governments on their degree of cooperation on these issues, and the near-term outcome of this is uncertain. International cooperation here is both necessary and achievable: There are clear precedents in treaties on nuclear weapons, telecommunications, and outer space that have been cited by both governments and AI industry players as templates for an international agreement on AI. Even with fits and starts, I see this as the eventual path forward to controlling existential risks.

In the US, I would not be surprised if we see some new AI regulations pass with bipartisan support in 2027 or 2028 given the high degree of public concern about both practical and existential risks. I expect these will continue to be extensions of what the industry has been pursuing through self-regulation, what AI safety institutes have advised, what government agencies have already been implementing in partnership with the industry, and what most groups agree need to be addressed as existential risks.

This does not mean that all practical AI safety concerns are completely resolved. Far from it. In parallel to the details of how US regulations and international cooperation play out, I think investors need to monitor how other industries are progressing in establishing their own AI safety controls when they deploy AI, since many practical issues occur at this level and regulatory responses are even more in their infancy. Public concern over data centers is another area to watch, especially in the US, where worries about AI safety and local community impacts are likely to persist through and beyond the US midterm elections. I see paths to resolution for both aspects, but reaching the desired destination will not be easy and I think investors will need to monitor developments here alongside progress on regulation and international cooperation.

Overall, I’d stress that the AI industry is still in its relative infancy. We should expect more advancements in AI and more choppiness in regulation, AI adoption, and the data center buildout each time a new advancement occurs. As long as the path to resolving AI safety and related concerns stays clear and policymakers and the industry can sufficiently keep pace, these moments of new advancement and renewed scrutiny may provide opportunities for active AI investors.

In the US and other developed economies, we're seeing that people — who are also voters — are becoming much more concerned about the impact of AI in their own backyards, whether it's data center construction, the price of electricity, issues of child safety or national security, or AI’s impact across the economy as it grows.

With something that becomes as pervasive as AI is supposed to get, regulation is almost inevitable. The question is when, how much, and in what form. It takes time for legislation to occur. Even if something needs to happen, it doesn't always happen right away. There is a significant risk that, as with other technological innovations, it will take an accident or other major incident to spur action.

Current US politics especially reflects this dynamic. The Republican Party has chosen to stress the pro-growth aspects of AI, and the president especially has pushed in that direction, arguing for staying the course given competition with China. The Democratic Party, meanwhile, has expressed greater concern about the technology’s risks.

With President Trump in charge, the focus has been on maintaining US leadership in AI while taking a lighter-touch approach to regulation. However, I expect that the 2028 US presidential election is going to see many candidates embrace AI regulation as part of their platforms. Following the election, we could even see bipartisan support for AI regulation.

Meantime, I'm watching regulation at the US state level more closely, as states often serve as precursors to federal action, experimenting on a smaller scale with policies that may eventually become templates for national regulation.

One thing that's been clear to me is that the pace of change in AI vastly exceeds governments’ ability to keep up through either legislative or broader regulatory issues.

Most policymakers recognize that AI is strategically important. Most governments acknowledge that AI is becoming a source of economic and geopolitical power. Recognition, though, is the easy part. Execution is where this all gets tricky.

Can governments create effective oversight systems? Can institutions build trustworthy deployment frameworks? Can companies absorb disruption and adapt fast enough? Can governance keep pace with capability?

For most of the last decade, the dominant investment question in AI was capability. Which model is strongest? Who has the most compute? Who wins the benchmark race?

Those questions still matter, but they're increasingly not the most important ones.

The emerging challenges are different. Who can verify increasingly powerful systems? Who can audit them? Who controls access? Who establishes standards? Who governs deployment? Who can safely integrate these systems into governments, corporations, militaries, and financial markets?

AI is no longer simply a technology race among companies. It's increasingly a strategic competition among states. The United States worries that slowing innovation could leave it vulnerable to Chinese advances. China worries about maintaining technological leadership while preserving political control and social stability. Both governments increasingly treat AI as critical infrastructure. That reality creates a tension that neither side can escape: Moving too slowly carries strategic costs, but moving too quickly creates governance risks that could prove equally consequential.

The traditional AI safety argument says we should slow down because the risks are rising. The national security argument says we can't afford to slow down because strategic competitors won't.

Both arguments can be true simultaneously. And that's precisely why AI increasingly resembles strategic infrastructure rather than a commercial technology.

For investors, that suggests the next phase of AI may look very different from the initial one. The first was a race to build intelligence. The next chapter is likely to be a race to govern it. If that's true, then the most important investment opportunities may not be found solely in the companies creating frontier AI, but in the companies building the systems that allow society to trust, deploy, secure, scale, adapt, and sustain it.