Lawmakers face an impossible choice: regulate AI too early and risk stifling innovation, or wait too long and let harmful effects spiral out of control. A new framework from Harvard Law School suggests a third path: pass laws now that automatically adapt as AI advances, triggering different regulations depending on what actually happens.

What Are Adaptive AI Laws and How Would They Work?

Martin Skladany, an AI and intellectual property law professor at Penn State who has advised groups including Apple and the Uniform Law Commission, argues that legislators should borrow a concept from machine learning itself: decision trees. Instead of passing rigid rules, lawmakers could create contingent laws that specify in advance how regulations will change based on AI's actual effects on society.

The idea is straightforward but powerful. Politicians could pass one set of laws today that automatically activate if job losses from AI mount, triggering policies like supplemental unemployment benefits or universal wage programs. A completely different set of regulations would kick in if AI instead drives job growth, perhaps mandating improved sick leave or reducing corporate subsidies. Benefits and taxes could rise or fall using sliding scales tied to measurable outcomes like employment rates or income inequality.

"Lawmakers can use numerous underutilized legal tools, such as sliding scales and triggers, to craft adaptive AI laws that trigger higher levels of safety regulation the more powerful AI models become," explained Skladany in the Harvard Journal of Law and Technology.

Martin Skladany, AI and IP Law Professor at Penn State

The framework could extend far beyond employment. Adaptive laws could mandate stricter accuracy standards for medical AI tools that diagnose patients compared to creative research applications where scientists can filter out hallucinations without patient harm. As AI capabilities evolve, the regulations would automatically adjust without requiring new legislation each time.

Why Can't Lawmakers Just Wait and See What Happens?

The core problem is timing. If politicians pass AI regulations too early, they risk inadvertently discouraging scientific discovery and innovation. But if they wait too long, harmful effects can already be entrenched, creating powerful interests resistant to change. Once problems gain momentum, they become exponentially harder to address because more actors are affected, dependencies multiply, and solving them requires vastly more resources.

Current AI leaders, including Geoffrey Hinton and Yoshua Bengio (two of the three "godfathers" of AI), have publicly warned about unaddressed dangers. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have both repeatedly called for regulation to improve AI safety and address potential job losses. Yet politicians remain hesitant to legislate a field where even the experts building AI models admit they don't fully understand how their systems work.

How to Implement Adaptive AI Regulation: Key Strategies

• Proactive Problem Prevention: Lawmakers can act now to avoid or lessen future problems before they spiral out of control, shifting from reactive crisis management to forward-thinking policy design.
• Comprehensive Coverage Without Sacrificing Flexibility: Adaptive laws can address multiple areas of AI simultaneously, from how models operate to their societal effects, while remaining amendable as technology and social conditions change.
• Stable Regulatory Clarity for AI Labs: By informing AI companies in advance which regulations will automatically trigger based on specific outcomes, the framework reduces uncertainty and encourages responsible development rather than reckless competition.

Skladany argues that adaptive AI laws would also encourage deeper thinking among policymakers. Rather than guessing which path AI development might take, politicians would systematically consider various scenarios and their potential harms or benefits. This structured approach could lead to more thoughtful policies than the current ad hoc approach.

The regulatory stability matters economically too. Economists have long documented that predictable regulation creates better outcomes for businesses and society. Adaptive AI laws could smooth the transition from today's virtually unregulated AI landscape to a more structured one by making clear in advance when regulations would activate. This means laws could be passed now with few immediately taking effect, giving the industry time to adjust.

Without any regulation, society faces what Skladany calls a "caveat societas" scenario, where the public bears all the risk. Traditional consumer protection concepts like "buyer beware" and "seller beware" became obsolete because they led to dangerous toys, cars, and medicines. Regulation brought safety and stability. Yet the current AI landscape, with virtually no regulation, is arguably more dangerous because the stakes are higher and the technology more powerful.

The challenge ahead is whether lawmakers will embrace this adaptive framework before AI's effects become impossible to manage. The window for proactive regulation is narrowing, and the stakes for getting it right have never been higher.