Skip to content

California advances twenty-six bills on AI and social media

The state where most of the labs are based is legislating at a pace that neither Washington nor Brussels can match. What is decided there applies in practice everywhere.

Advertisement
The essentials in 30 seconds ⚡
California has advanced twenty-six bills on AI and social media. It is the state where most of the labs and platforms concerned are based. What gets decided there ends up applying well beyond its borders, for a purely practical reason.

We covered the entry into force of the California law on content watermarking. Here is the next wave, and the mechanism is worth explaining.

Why this state carries so much weight

Three reasons combine, and the third is the most important.

Concentration. Most of the companies targeted are headquartered there. A California obligation therefore applies directly to those building these systems.

Market size. No company gives up this market to avoid a regulatory constraint.

The cost of differentiation. This is the decisive mechanism. Developing two versions of a product — one compliant for California, another elsewhere — costs more than applying the rule everywhere. Companies therefore align their global product on the most demanding jurisdiction.

This is exactly the mechanism that gave the European data regulation global reach. A local rule becomes a de facto standard when complying everywhere costs less than segmenting.

The consequence for a European reader 🌍
This means that part of what actually governs the products you use is decided neither in Brussels nor in Paris, but in Sacramento. This is not a judgement, it is an observation on how tech regulation works in practice: it is produced by the jurisdiction that combines a significant market and an ability to legislate quickly. And on that second point, a US state legislature moves considerably faster than a federal or European process.

What these bills generally cover

Without going into the detail of each one, the recurring themes of this legislative wave overlap with what we follow.

Protection of minors, a topic on which we covered ChatGPT's teen mode and China's ban on emotional companions for minors.

Transparency, in line with content watermarking and the European declaration requirement.

Automated decisions concerning individuals, a topic we addressed regarding recruitment.

The tension with the federal level

One point structuring this issue should be noted. The US federal government has worked on an evaluation framework whose content has not been published, and part of the industry is arguing for a single national rule that would replace state legislation.

The argument is reasonable: a company facing fifty different regulations bears a considerable burden. The counter-argument is equally so: a federal rule negotiated with the industry risks being less demanding than those of the strictest states.

This is a classic division-of-powers debate, and its outcome will largely determine the real level of constraint in the years to come.

What to take away

To understand what actually governs these technologies, you need to look at three places: Brussels for the most developed framework, Sacramento for the fastest-moving one, and Beijing for the one that applies most directly.

What is striking is that none of these three is the US federal level, even though that is where the companies concerned are based. A vacuum at the centre, filled by the peripheries: that is an unstable configuration, and it explains why the companies themselves end up calling for a rule, if only to know which one to apply.

Advertisement