MUNICH — Shortly after wandering into the Munich Urban Colab this summer, I found myself trying to follow an intense conversation in German about the city’s new experiment with autonomous buses. A local official, clocking my curious puzzlement, offered an aside in English: “We are asking ourselves: ‘What does it mean when the law is an experiment too?’”
No question may be more essential to governance in our era of dangerous uncertainty.
Global warming is accelerating faster than global treaties on mitigation can be negotiated. Climate catastrophes occur with a sudden ferocity beyond what current adaptation laws anticipate. Biological threats surface and kill before public health responses can be formulated. And AI is evolving faster and more wildly than regulators or lawmakers, or even its designers, expect.
All this dizzying unpredictability is undermining traditional legal structures. The threat is not just to the rule of law, but to the very nature of law, at least as humans have conceived it.
Our societies have long considered the law a slow and steady force, “the essential foundation of stability and order” in the words of J. William Fulbright. To that end, law is backward-looking — “the government of the living by the dead,” according to Oliver Wendell Holmes. The lawmaking process in republics and democracies is age-old: observe a problem, study its particularities, debate solutions and then carve out a rule that will endure for years or decades.
But what happens when the future arrives faster than your regional regulator or local parliament can make a rule?
The emerging answer is experimental law or, in Germany, where Munich is a hotbed, Experimentierklauseln. “Experimental clauses” refer to sections and parts of laws that are made flexible, often for set time periods, to allow for real-time adaptation or alteration in response to changing threats, climate or technology.
The shift toward such experimentation has begun, albeit slowly, in only certain areas of the law. Contract law and administrative law, especially where they touch on technology or other advanced sectors of the economy, are becoming more fluid.
Contracts, once used to protect against all contingencies, are now written more openly — explicitly anticipating new knowledge and adjustments. Among startups and innovation-oriented companies, written agreements are less likely to specify payments for specific work at specific times and more likely to establish processes for collaboration and shared decision-making.
For example, start-ups use Agile contracts with vendors that, instead of pledging a final digital product at a price, set parameters of a creative process involving a test phase, iterative feedback loops and product benchmarks. Biotech startups rely on a “relational contract” with manufacturers to create a collaboration framework that includes regular meetings, shared lab notes and a commitment to use best efforts to scale a prototype.
Meanwhile, smart contracts have become commonplace in supply chains to allow for real-time changes in inventory levels. Union contracts in digital workplaces often incorporate adaptive provisions — “continuous bargaining” in 21st-century labor parlance — with guidelines for how workers participate across the lifecycles of systems and products.
Modern tech procurement uses what’s called “Dynamic Service Level Agreements” that feature automatic updates, scalable license adjustments, and continuous remote monitoring and interventions by seller and buyer. “Data Processing Agreements” within cloud contracts dynamically adjust data handling and security protocols as privacy regulations and digital risks shift. In some cases, AI agents have written contracts in code rather than in English. Put simply, some contracts now write themselves.
Similarly, regulatory agencies and public administrations have moved away from the time-consuming traditional rulemaking process, since rules quickly become outdated in a world so strange and fast-changing. Instead, agencies and administrative judges, particularly in advanced economies, are more likely to offer “guidance” — advisories, memos, letters or protocols that are regularly tweaked or edited in response to updated knowledge and changing circumstances.
These shifts point toward an emerging and quite radical recognition, especially in law, that extreme uncertainty is not a problem to be solved, but a reality and even an ally. The fact that the future is too hard for anyone to know is leveling. The regulator, the regulated and the everyday person all know the same thing: almost nothing.
The act of lawmaking then becomes a great collaboration between governments and experts and firms and workers and the regular people most affected. And that collaboration is a search, never-ending, for improvements and innovations and solutions.
Such collaboration turns lawmaking as we know it on its head. Implementation is required first to understand what the policy should be. In this new paradigm, resilience is grounded in fragility. Resilience does not come from buffering organizations against known risks, but by designing them to show quickly when they are broken, so they can be reconfigured. It is a kind of deliberately induced and carefully controlled institutional fragility. Strain is immediately apparent, triggering investigation and countermeasures well before it ramifies into catastrophe.
“Experimental clauses allow for real-time adaptation in response to changing threats, climate or technology.”
Climate Changes The Law
The stakes of achieving this new, more open mentality are expressed perhaps most clearly in our response to global warming. In the 2022 book, “Fixing the Climate: Strategies for an Uncertain World,” Columbia Law School’s Charles F. Sabel and UC San Diego’s David G. Victor argue that decades of top-down global climate diplomacy — from the Kyoto Protocol to the Paris Agreement — have failed because they rely on inflexible global treaties.
Treaties are among the oldest forms of law. But climate treaties are outdated the moment they are signed, because the climate is changing too quickly. And they don’t reflect the reality that climate change impacts — and opportunities to mitigate and adapt to climate change — are very different depending on location and context. (Sea level rise is not an issue in Mongolia; glacial melting does not plague Phoenix). The only viable way to make law and government is thus through relentless, highly structured and broadly collaborative local experimentation. This model, they noted, is neither “top down” nor “bottom up” but both at the same time.
“Our central aim is to reorient our current climate change regime away from failed efforts based on ex ante global consensus, and toward a system anchored in local and sectoral experimentalism and learning,” they wrote.
This local experimentalist mindset is emerging especially in the metro regions of countries that have advanced economies and histories of collaborative lawmaking and regulation.
Two places that best fit that description are Singapore, which was built on deep cooperation between government technocrats and industry, and Germany’s bigger cities.
Experimental law first emerged in West Germany during the 1950s as a pragmatic mechanism for the Länder, or states, to exploit their autonomy under the 1949 Basic Law. Faced with ideological divisions and a lack of national precedent for building a post-authoritarian society, municipal and state agencies used “trial clauses” to break legacy rules on welfare, housing and local administrative reorganizations without setting precedents or locking local governments into long-term mandates. The first official Experimentierklauseln allowed public schools to violate rigid curricular and education standards and test pedagogies, grading systems and models of comprehensive schools.
Experimental clauses fell out of fashion in the late 20th century but were revived for new challenges in the 21st century. In 2008, the Upper Saxony town of Bohmte, Germany, used an experimental clause to do something otherwise illegal: eradicate traffic lanes, street lines, traffic lights and stop signs to combat chronic speeding, utilizing a Dutch idea called shared space.
Over the last decade, climate change and advances in technology and transportation inspired so much local experimentalism that the German Federal Ministry for Economic Affairs and Climate Action issued an official “Guide for Formulating Experimentation Clauses” in 2024. The guide runs 34 pages but contains three main rules: All experimental clauses must have an expiration date, a clear geographical or sectoral boundary, and rigorous evaluation.
Ask the federal ministry staff for a signature example, and you’ll hear about a Hamburg hospital that couldn’t get cancer tissue samples to a laboratory in time because of city traffic. The hospital wanted to use a drone instead, but German aviation regulations strictly forbid drones from flying “beyond the visual line of sight” of their operators over settled urban areas.
Instead of waiting years for federal airspace reform, Hamburg relied on experimental clauses to deviate from the federal Air Traffic Act for 18 months. During that time, the city-state established the U-Space Sandbox, a 30-square-kilometer experimental airspace over the Port of Hamburg.
Officials of the local and national governments worked with two German companies, Droniq and DFS Deutsche Flugsicheruntegration, to test the integration of drones with manned aviation. The experimentation eventually produced guidance and rules for a drone air traffic system, which are being incorporated into drafts of Germany’s federal U-Space Act.
“Climate treaties are outdated the moment they are signed, because the climate is changing too quickly.”
Living With Experimental Law
In Munich, I spent an evening inside one product of experimental law: an unusual new house at Goerzerstrasse 128, in the southeast neighborhood of Giesing. Munich, like so many cities, suffers from high housing prices and rents. The city held a competition for alternative solutions, and a self-organized housing syndicate won. It then used experimental clauses to deviate from strict German building codes.
The building was constructed with simple wood and cheap corrugated iron — what residents call “Favela-style” — and operates internally as a massive “shed flat.” The design is energy-efficient, and the roof captures rainwater in an underground cistern and feeds it directly back into the local groundwater table to combat drought.
The cost exceeded 3 million euros, drawn from public subsidies and private loans. But that’s cheap, given its scale. For now, 14 people — 13 adults and a six-year-old — share the open architecture, with frequent visitors. (A rock band was just leaving as I arrived.) The building is designed so it could be divided into at least nine separate apartments. While one-bedroom apartments in Munich rent for 1,700 euros a month, rent at Gorzer 128 is about 460 euros per person.
Gorzer 128 and other experiments helped inspire a national experimental clause called “Construction Turbo” (Section 246e of the German Federal Building Code), which allows municipalities to bypass existing planning standards until the end of 2030. The only unhappiness has come from neighbors who don’t like the building’s appearance (a jumble of metal and wood, quite different than the single-family craftsman homes on the street) and have been slow to engage with residents. Experimental law offers no cure for NIMBYism, at least not yet.
Beyond Gorzer 128, Munich’s experimental laws are most associated with efforts to transform streetscapes. A signal success is on the Kolumbusstraße, which a few years ago was a standard urban artery, lined with 120 concrete parking slots and choked with traffic.
Oliver May-Beckmann, who leads the MCube mobility cluster at the Technical University of Munich (TUM), says his group used an experimental clause to close the street to traffic for six months to create a Sommerstraße (Summer Street). Some drivers and residents complained to local officials. But because the project was a temporary and science-driven Reallabor, or real-world laboratory, politicians could push responsibility onto May-Beckmann and the scientists. “They’d say, ‘It’s just an experiment, let’s wait for the data,” he recalls.
For half a year, Kolumbusstraße became a green space where children could play, residents could spend time and new forms of mobility were tested. The transformation was deliberately temporary, and after six months, the street returned to its previous state. But the evaluation produced a striking result: 66% of residents supported making the changes permanent. Whether and how to turn the experiment into a lasting transformation is now a decision for the City of Munich.
MCube also used an experimental clause to test a car-sharing strategy in Nymphenburg, an affluent neighborhood that houses Munich’s botanical garden. For 20 years, Germany’s car industry — BMW, Mercedes, Sixt — had been testing different forms of free-floating vehicle-sharing to try to cut emissions, but it just put more cars on the street because people used shared vehicles on top of their own. In MCube’s test, 10 local families gave up their private cars for 18 months, and MCube bought five electric cars, plugged into charging stations in an underground garage, for the families to share. It was a closed group — the sharing was just between the 10 households, and you knew who had the car before you.
The families communicated with each other and shared the cars well. MCube also discovered that when people know a car is always nearby, they were willing to pay significantly more in regular payments for access than they would pay with one-off models. MCube is now working with both BMW and Sixt, the Bavaria-based rental car company, to spread the model around Germany. The experimental clause is still in effect and is being used to test alternative car-sharing models in three other Munich neighborhoods.
Experimental clauses offer flexibility but demand deep collaboration with business, governments and everyday people, May-Beckmann says. Take MINGA, a new project that plans to put autonomous buses on Munich’s streets this fall. (Minga is both the name for Munich in the Bavarian dialect and an acronym for Münchens Automatisierter Integrations-Nahverkehr mit Gesellschaftlicher Akzeptanz, or “Munich Automated Public Transport with Ridepooling, Solo Bus and Bus Platoons.”
The project, which uses driverless vehicles without a set schedule or route, is enabled by experimental clauses derived from 2021 amendments to Section 1i of the German Road Traffic Act. The German government is providing 13 million euros in funds. MAN, a German commercial vehicle manufacturer, built the buses. Technical University of Munich scientists have worked on the software, the public transport administration is preparing to run the buses, and the city of Munich plans to operate the service for two years under experimental clauses and then evaluate it. Finally, it requires people to sign up, primarily to bridge the “first-mile/last-mile” gap of getting people to public transit hubs.
Experimental clauses, at their best, create space and permission for systems integration. That’s nearly impossible in conventional governance, where decisions are made sequentially, and services and programs are added one by one.
“Our goal is transferring knowledge into prototypes and bringing those into real life,” says May-Beckmann. “It’s crucial in Europe that mobility innovation isn’t just about a new car; it’s about a system. How do you integrate autonomous driving, city development, climate change and energy? We get funding to have industry and public administration work for us as a system.”
“All experimental clauses must have an expiration date, a clear geographical or sectoral boundary, and rigorous evaluation.”
Experiments In Sunshine & Medicine
The desire to experiment with law is not exclusively German. Barcelona placed the 22@ Innovation District under experimental law to test drones. In the UK, Bethesda and other Welsh cities have employed experimental clauses for “Sunshine Tariffs” to encourage renewable energy use. South Korea carved out an experimental exception in its transportation laws to install a hydrogen-powered mobility system in Gyeongsangbuk-do. In Japan, the experimental mantle is carried by the “Future Design” movement, where citizens in towns like Yahaba wear differently colored robes to design infrastructure plans for future generations.
If democracy doesn’t have to be part of the legal equation, then East Asia might be the world’s leader. Singapore, the world leader in temporary regulations, has built a financial powerhouse by permitting banking and fintech experiments for decades. More recently, it has been applying experimentalism to medicine. An ongoing Health Sciences Authority experiment allows low- to medium-risk AI medical software to be used in public hospitals before formal registration. If the AI performs well for 12 months under clinician oversight, it can get an accelerated pathway to permanent approval.
More broadly, China and India have survived and prospered during rapid transitions by making experimentation a foundational principle of governance, often at the municipal or provincial levels.
In his 2024 book “Experimentalist Constitutions,” Peking University legal scholar Yueduan Wang details the Chinese government’s use of “experimental zones” to deviate from national law. In several areas, including green energy and mobility, municipalities have explicit legislative permission to try things. For example, the Shenzhen People’s Congress passed the 2025-2026 “Decision on Amending the Regulations of Shenzhen Special Economic Zone on Intelligent Connected Vehicles,” which allows fully driverless commercial taxis to bypass national safety standards.
India has embedded what it calls “regulatory sandbox” clauses directly into legislative updates. The Telecommunication Act of 2023 explicitly created a live testing environment that exempts companies from standard licensing fees to test 6G internet. Gujarat, where Narendra Modi served as chief before becoming national prime minister, is an epicenter of experimentation. It houses India’s first operational “smart city,” Gujarat International Finance Tec-City (GIFT City), between Ahmedabad and Gandhinagar, to test flexible financial regulations. Many of its startups use an “interoperable sandbox” to temporarily bypass rules on corporate governance, liquidity, capital minimums, reporting timelines and marketing as they test hybrid fintech products.
Experimental Enough For AI?
Despite successes in fintech, mobility and housing, it is unclear whether experimental law will be fast or flexible enough for the ethereal, borderless realm of artificial intelligence.
In Germany, companies working with AI complain that experimental law remained too slow and earthbound to keep up. Ironically, they blame the European Union’s attempts to “lead” on experimental law and AI for the failure.
In 2024, the EU embraced experimental law with its landmark AI Act. Article 57 of the AI Act included an explicit mandate for experimental law — every member state was required to establish at least one national “AI regulatory sandbox” by August 2026. But the mandate made the program politically divisive within several countries. In a dozen member states, including Germany and France, governments couldn’t even agree on which national agency or authority would establish their sandbox. The EU acknowledged the dysfunction this summer, pushing the deadline for a national AI mandate back a year to August 2027, via the AI Omnibus.
Few in Europe’s AI policymaking community expect to meet that deadline. The trouble goes beyond the EU’s slow bureaucracy. It’s hard to employ experimental law when you’re terrified of real impacts and results. Notably, the AI Act doesn’t require experimental deployment of AI regulation in the real world. As a result, those nations that are moving toward an AI regulatory sandbox are limiting experiments to theoretical exercises or purely legal assessments. But it’s impossible to create experimental laws for AI in the real world if you won’t experiment with AI in the real world.
In Munich, Till Klein, head of regulation at the non-profit “appliedAI Institute for Europe,” says that regulatory sandboxes that do exist for AI simply have too many restrictions to be useful. “Policymakers act like sandboxes will fix all the new rules,” says Klein. “But a sandbox is more like a sniper rifle than a shotgun; your experiment has to be precise for it to be meaningful.”
His colleague Demian Niemeyer, a lawyer and senior AI regulatory expert at appliedAI, sees the choke point as cultural. “People trained in law here are trained in a ‘Bismarck mindset,’” he says. “They are incredibly risk-averse. If you want to make something better and understand risks, you need a trial-and-error mindset.”
The paradox is that fear of applying experimental law to AI has left AI running free, well beyond law and regulation. Another fear also hamstrings experimental law: fear of regulatory capture.
That’s understandable, since effective experimental law requires regulators to collaborate closely with companies and stakeholders. Experimental law challenges the conception of regulation and law as walls between the watcher and the watched.
“The paradox is that fear of applying experimental law to AI has left AI running free, well beyond law and regulation.”
In 2024 and 2025, fear of capture stirred controversy and delayed the drafting of Germany’s federal Reallabore-Gesetz (Regulatory Sandboxes) Act, an attempt to go beyond experimental clauses on particular subjects and create a broader framework for experimental law.
The Social Democrats, left-wing groups and some scholars argue that broad authorization of experimental clauses risks turning into backdoor-lobbying channels for Big Tech and powerful corporations seeking exemptions from labor, safety and privacy laws. Most of the objections were AI-focused.
In response, the proposed law’s backers incorporated regulatorisches Lernen (regulatory learning) as a sandbox objective. When a project is approved under an experimental clause, the outcomes cannot remain hidden. The law requires continuous evaluation of the testing, and the resulting insights must be reported transparently. There are limited exclusions for genuine intellectual property.
The law also seeks to avoid narrow regulatory discussions between a company and a government agency via a “Regulatory Sandboxes Innovation Portal.” The portal is supposed to facilitate broad co-creation of law by connecting stakeholders from business, academia, public administration and civil society.
The legislation, renamed the Bundeserprobungsgesetz (Federal Proving Act), passed the Bundestag in late June. A second legislative body, the Bundesrat, would formally enact it as law in September. At the Munich Urban Colab, the immediate reaction was uncertainty. People would have to experiment with the new law and see.
