Outcry from the economy – The fear of the AI ​​​​law

The signatories criticize the Brussels “AI ACT” especially with regard to rules for generative AI. These are systems that can automatically produce text and images, for example. Generative AI developers would face “disproportionate compliance costs and liability risks,” they write.

Why is? Regulation of AI – that is the AI ​​Act

The European Union (EU) already published a strategy for dealing with artificial intelligence in 2019. It sets guidelines to strengthen research and business in this area, while ensuring security and protecting fundamental rights.

The EU wants to set the legal framework with the AI ​​Act. The process is well advanced, and the Commission, Parliament and Council want to come to a final agreement by autumn. So the cornerstones have long been clear. The law is intended to divide AI applications into risk classes: “minimal”, “limited”, “high” and “unacceptable”.

The greater the potential consequences, the more providers need to do. Some systems are even banned from the outset – such as surveillance using real-time biometric facial recognition.

The bureaucratic hurdles increase from risk level to risk level. The focus of the AI ​​Act is on high-risk applications, which include robotic surgery, personnel selection systems or critical infrastructure.

>> Read also: “Generative AI will never be able to replace SAP,” says Dax CEO Christian Klein

Anyone who wants to get involved here must, among other things, introduce risk management, inform users “clearly and appropriately” and submit technical documentation with detailed information on the data used.

Why is that important?

Artificial intelligence needs rules, that is undisputed. Since the AI ​​chat program was launched in November 2022, new applications have been developing at a rapid pace. Some even warn of end-time scenarios in which the AI ​​becomes independent and turns against humans.

In the AI ​​industry itself, some players are calling for a halt to development. The senders of the open letter also see an “undeniable need for appropriate regulation”.

The use of the technology already entails risks: disinformation campaigns that use artificially generated photos and videos. Control systems used in critical infrastructure such as the power grid. Or algorithms that are supposed to recognize social fraud, but discriminate against minorities. This has already happened in the Netherlands.

The dangers lie in the way the technology itself works. Artificial intelligence must first be trained. For this you need a lot of data as learning material. And that can include human prejudices. Sometimes people don’t see it at all. And later it is difficult to understand how the AI ​​arrives at its conclusion. Their algorithm is a black box. Generative systems that can generate text, photos or program code also enable deceptively real counterfeits.

What are points of contention?

In the EU there is hope for the “Brussels effect”: As with data protection, other countries should take up the legislation on artificial intelligence. When that happens, European ideas will become an international standard.

However, many business representatives fear much more that the AI ​​Act will become a locational disadvantage. The senders of the letter write: “Such regulation will in all probability lead to highly innovative companies relocating their activities to non-European countries.”

Investors could therefore shy away from investing in the development of European AI models. “The result would be a critical productivity gap between the two sides of the Atlantic.” The signatories to the letter would like to see changes on two points:

• First call for the “building of a transatlantic framework”: Many important US actors have made similar proposals as the European ones. The European Commission should use this to create fair competitive conditions on both sides of the Atlantic.

• Secondly are pushing for more flexible rules for generative AI. “Wanting to enshrine the regulation of generative AI in a law and to proceed according to a rigid compliance logic” is “an approach that is as bureaucratic as it is inefficient”. Instead, the EU should initially only lay down general principles. The implementation should then be entrusted to a body of experts that is able to “continuously adapt these principles to the rapid pace of development of this technology and the emerging concrete risks”.

Patrick Glauner, Professor of Artificial Intelligence at the Technical University of Deggendorf, understands the criticism. “The requirements for high-risk systems are difficult for many companies to meet,” he says.

In his view, every start-up and every medium-sized company should create their own AI compliance department if they use artificial intelligence. He thinks it is possible “that certain innovations will no longer take place in Europe” – for fear of bureaucracy and penalties.

>> Read also: Which ministries are already using AI – in surprisingly concrete projects

And that would probably affect a lot of companies. In a survey of publicly available AI systems, the Applied AI initiative classified 18 percent as high and 42 percent as low risk.

For 40 percent, the assessment is unclear – and that could also be problematic. “An unclear classification slows down investments and innovations,” explained the organization, which belongs to the start-up center of the Technical University of Munich (TUM) and wants to promote the application of the technology.

How big is the chance that Brussels will change something else in the AI ​​Act?

The open letter from the founders and managers is late – the EU Commission published a first draft of the AI ​​Act as early as April 2021. “I have the impression that for a long time there was very little from the companies,” says Glauner, who has appeared several times as an expert in the Bundestag. He attributes this to ignorance on the one hand, and deliberate political restraint on the other: “If the economy had positioned itself like this a few months ago, it would have been much easier to exert influence.”