Germany ‘To Approve’ EU’s AI Act

Germany is set to give its approval to the EU’s AI Act after the country’s minister for digital issues said a compromise had been reached.

The European Parliament and EU member states under the European Council in December reached a provisional agreement on the rules for artificial intelligence, making the EU one of the world’s pioneers in regulating the area.

The deal still has to be approved by member states and the European Parliament.

German digital minister told Reuters on Tuesday he had pushed for more innovation-friendly rules and had achieved improvements for small and medium-sized businesses.

AI competitiveness

The changes avoid disproportionate requirements, he said, without giving details.

“Without the use of artificial intelligence, there will be no competitiveness in the future,” Wissing said.

“The wrangling over the German position on the AI Act came to an end today with an acceptable compromise.”

He added that the “negotiated compromise lays the foundations for the development of trustworthy AI”.

Government oversight

Wissing’s Free Democratic Party is considered pro-business and campaigns to protect civil liberties.

The FDP is part of Germany’s ruling coalition with the Social Democrats and the Greens.

December’s agreement followed a stand-off that hinged on the treatment of so-called foundation models, the most powerful forms of AI.

France, Germany and Italy had reportedly campaigned for foundation models to be left to self-regulate, in order to ensure EU companies would not be competitively hampered by excessive red tape, while the European Parliament had argued the technology was too important to be governed by companies on their own initiative.

Tiered system

Turmoil last November at AI pioneer OpenAI, which saw chief executive Sam Altman sacked and then re-hired a few days later, fuelled the arguments of those who said companies should not be left to self-regulate.

“We cannot rely on voluntary agreements brokered by visionary leaders,” said Brando Benifei, one of the two European Parliament negotiators, at the time.

The December agreement reportedly puts into place a risk-based tiered system with the top level of regulation applying to the most powerful AI models, defined as those that require the most computing power to train.