A.I. Regulation Is Coming Soon. Here’s What the Future May Hold
If you want to know how the global regulation of artificial intelligence might shape up in the coming years, best look to Berlin.
Last year Angela Merkel’s government tasked a new Data Ethics Commission with producing recommendations for rules around algorithms and A.I. The group’s report landed Wednesday, packed with ideas for guiding the development of this new technology in a way that protects people from exploitation.
History tells us that German ideas around data tend to make their way onto the international stage. That’s what happened with online privacy—German rules fed into the EU-level General Data Protection Regulation (GDPR), which is influencing global policy because of the EU’s economic weight. And it seems likely to happen here, too. Indeed, the Data Ethics Commission specifically calls for a new, EU-wide set of A.I. rules, based on its recommendations.
So, what do those recommendations look like? In a word: tough.
The commission insisted that algorithmic systems should be designed safely, to respect people’s rights and freedoms, protect democracy, be secure, and avoid bias and discrimination.
It said systems presenting a significant risk of harm, such as those that show different people different prices based on their profiles, should in some cases require licensing. And systems with an “untenable potential for harm”—killer robots, for example—should be banned outright.
The group—whose members work in academia, industry and regulation—also called for a mandatory labeling scheme that would apply to algorithmic systems that pose any potential threat to people’s rights, and said people affected by algorithmic decisions should be able to get “meaningful information” about how those decisions were reached.
It also called for an update to liability rules, to make sure companies can be punished for rights violations and bad decisions made by algorithms that would otherwise be made by human employees.
“The Data Ethics Commission recommends that measures be taken against ethically indefensible uses of data,” it wrote. “Examples of these uses include total surveillance, profiling that poses a threat to personal integrity, the targeted exploitation of vulnerabilities, addictive designs and dark patterns, methods of influencing political elections that are incompatible with the principle of democracy, vendor lock-in and systematic consumer detriment, and many practices that involve trading in personal data.”
Big Tech’s lobbyists are not happy.
“The German Data Ethics Commission’s recommendations on A.I. send a worrying signal to businesses that they risk adopting A.I. at their own peril,” said Eline Chivot, a senior policy analyst at the Brussels office of the Center for Data Innovation—a think tank that reliably defends the tech industry against heavy regulation (the center does not disclose its funding, but is affiliated with the Information Technology and Innovation Foundation, which has taken money from the likes of Google and IBM).
“If Germany’s guidelines were to inspire the EU’s forthcoming A.I. legislation, the EU will indeed manage to set a global standard—a blueprint on what to do to fail in the digital economy,” Chivot wrote in a statement.
However, the Data Ethics Commission’s recommendations are by no means all business-unfriendly. For example, it said that when companies anonymize or “pseudonymize” people’s data, they should then benefit from legal certainty that they are protecting those people from the violation of their data-protection rights.
On the other hand, the commission also said the de-anonymization of personal data should become a criminal offense.
Businesses should be cautious about deploying A.I., one of the commission’s members, Marit Hansen—the data protection authority for the German state of Schleswig-Holstein—told Fortune. However, she said, the recommendations should not hamper innovation.
“Regulation is helpful and also good for companies because they know what guarantees to demand from their suppliers,” Hansen said, citing fields such as medicine and autonomous vehicles as examples where guarantees are needed.
The recommendations mostly came as music to the ears of civil society activists who have long been calling for tighter algorithmic regulation—albeit with some reservations. AlgorithmWatch, a German non-profit, said it welcomed some of the Data Ethics Commission’s ideas but found some aspects “blurry.”
“Exactly how an ‘overarching model’ to sort algorithms by their potential harm should be developed will raise new questions,” AlgorithmWatch noted. “Existing systems number in the thousands and several of them might fall between categories.”
However, it also pointed out that incoming European Commission President Ursula von der Leyen wants to propose new EU-wide A.I. regulations by March, and the fact that she is German suggests the recommendations “will find their way to the European level in the near future.”
Germany recently joined forces with France and Japan to work on “human-centered” A.I. at the research level. Some saw the move as marking a split between those countries and the U.S. and China, where A.I. is arguably being developed with fewer ethical considerations.
The U.S. and China are where the majority of A.I. development is taking place, so this ethically-focused alliance could represent the beginnings of a counterweight—and this week’s German recommendations could provide a first glimpse at their rulebook.
More must-read stories from Fortune:
—Green, nuclear, and crowdfunded: One startup’s unconventional route to building a novel reactor
—Faraday Future’s founder has filed for bankruptcy, but the Tesla rival says it’s not dead yet
—Norway is a green leader. It’s also drilling more oil wells than ever
—Plant-based burgers may be on the rise, but meat consumption is higher than ever
—China’s economic growth slows to a 30-year low. But is it the U.S. trade war?
Catch up with Data Sheet, Fortune’s daily digest on the business of tech.