Long-term artificial intelligence law: new controversy over ChatGPT rules

DThe call for strict regulations was not long in coming when ChatGPT created a buzz around generative artificial intelligence (AI) in the spring. Even Silicon Valley warned of the risk of human “extinction” due to such AI. The European Parliament quickly introduced and adopted in June strict requirements for AI developers who are creative in creating text, images and videos. But now Germany, France and Italy are opposed to each other. In a joint document they speak out against legal regulation.

“We are opposed to the introduction of untested laws and therefore propose to rely initially on mandatory self-regulation through a code of conduct,” said an internal document circulated on Sunday. It can be obtained from FAZ. The Code of Conduct is intended to oblige generative AI manufacturers to create a kind of insert (“model cards”) for their AI. This is intended to ensure a minimum level of transparency and security in accordance with the AI ​​principles of the Hiroshima Process initiated by the G7 countries.

Counterproposal: AI Package Insert

Leaflets must contain “all necessary information to enable one to understand how the model works, its capabilities and limitations.” discrimination and safety testing conducted by independent third parties (“Red Teams”). The document does not directly provide for penalties for violation of requirements. It only says that their introduction may be considered in case of repeated violations.

With the help of a catalog of demands, three states want to influence the ongoing negotiations on the planned EU law on artificial intelligence. Negotiators from the European Parliament, the EU Commission and the Council of Ministers have been trying to find a compromise for several weeks. The Artificial Intelligence Act aims to establish clear rules for the use of artificial intelligence for the first time. European lawmakers want to create a model for the whole world.

An agreement is due to be reached on December 6. Then the final “trilogue” of negotiators is planned. A newspaper from Germany, France and Italy will likely make this task much more difficult. EU parliament negotiators recently broke off talks and defiantly walked out of the chamber after Germany and France opposed generative AI regulation that had previously been seen as agreed among EU institutions. In parliament the new document was called a “declaration of war.” Other member states, including the Spanish presidency of the EU Council, which maintains de facto neutrality and negotiates in the Council of Ministers, are also pushing for strict EU rules on AI.

France as a driving force for soft regulation

The newspaper’s driving force is France, which wants to establish itself as an AI powerhouse rather than slow down its national AI champions such as Mistral. On Saturday, French Digital Minister Jean-Noël Barrault invited former Open AI CEO ChatGPT developer Sam Altman, now hired by Microsoft, to move to France “where we will intensify our efforts to bring artificial intelligence to serve the common good,” he wrote. Barro at short message service X, formerly Twitter. The EU Parliament also criticized the role of Cédric O, who, as a former digital minister, advocated for AI regulation but is now lobbying for it as a representative of Mistral.

There are differing views within the federal government (not least between the FDP and the Greens) about how carefully AI needs to be regulated. However, Economy Minister Robert Habeck (Greens) had already moved away from strict requirements before Heidelberg startup Aleph Alpha recently raised $500 million from investors to develop its language models to compete with ChatGPT. “We don’t want to regulate the technology, but the possible applications,” Habeck said at the artificial intelligence summit in Bletchley Park, England.

Federal Minister for Digital Technology Volker Wissing (FDP) praised the joint document with almost the same words: “We must regulate applications, not technologies, if we want to play in the world’s top AI league.” The problem is that it is unclear how it will be used. Therefore, unlike “conventional AI”, which is developed for a very specific application, its regulation must also occur independently of the application.