Patchwork

A patchwork is, in the literal sense, a quilt sewn together from many scraps of fabric. It holds together, but it has no continuous pattern, and the seams are visible everywhere. As an image, the word is used in the news when many different rules apply to one topic at the same time, instead of a single one. Anyone selling software or devices in several countries then has to comply with different regulations in each country. The term is always meant slightly negatively: a patchwork is never praised. It can also refer to technical systems made up of many cobbled-together individual parts.

Why companies fear this state of affairs

For a company, diversity in rules mainly costs money. A company that offers an AI assistant then has to check which regulation applies where. This requires lawyers in every market, and often different versions of its product as well. Large corporations can afford that. Small start-ups often cannot manage it and avoid certain countries entirely.

For users, this has an unpleasant consequence. New features appear in some regions months later, or not at all. Some AI features from Apple, Google, and Meta arrived in the European Union noticeably later than in the US. The companies each cited the legal situation as the reason. Whether that was merely a pretext is disputed.

There is, however, also an argument in favor of many individual rules. When regions try out different approaches, it becomes possible to see what works. Experts call this regulatory competition. Critics counter that companies then simply move to wherever the rules are most lenient.

How such a regulatory mess even comes about

Every state makes laws for its own territory. For an app that is accessible worldwide over the internet, this principle fits poorly. As long as there is no international treaty, many individual rules inevitably arise. A provider is then subject simultaneously to all the legal systems in which it has customers.

In the US, a second layer is added. There, the 50 states are also allowed to pass their own laws, as long as the federal government has not regulated the topic itself. Because there is so far no comprehensive federal law on artificial intelligence, states such as California, Colorado, and Texas have adopted their own regulations. They differ in details: what counts as a risky application, who has to disclose what, and above what company size. In the worst case, a provider must comply with all the variants.

The European Union takes the opposite approach. Its AI regulation, the AI Act, is meant to set the same rules for all member states and thus prevent a patchwork within Europe. Between Europe, the US, and China, however, the difference remains large. So the patchwork does not disappear, it just becomes coarser-meshed.

The word in headlines and in technology

In business news, you mostly encounter the term in quotes from managers and industry associations. A typical sentence is that a patchwork of state laws harms innovation. Such statements are often lobbying: whoever says them usually wants a uniform, ideally lenient federal law. So it is worth considering who is using the comparison.

Outside of politics, engineers use the word for systems that have grown organically. A bank whose software is assembled from dozens of old programs and stopgap fixes has a patchwork of IT. Related to this is the term legacy system, meaning outdated technology that nevertheless has to keep running. The difference: legacy refers to the age of a component, patchwork to the absence of an overall plan. Both make it harder to connect AI to them in a sensible way later on.