Do not water Europe’s AI rules to please Trump, EU MPs warns

MPs that help to form a significant AI act of the European Union, 27 members think that the AI ​​rules are irrigated by the pressure of US technology companies from US technology companies.

This I have an act Only a year ago, it was approved, but the rules for general AI models such as Openai’s GPT-4O will only come into force in August. The European Commission, the European Commission, the European Commission, the EU executive branch, has instructed the practice code by preparing an internship code for Big AI companies and how to comply with the legislation.

However, a group of European MPs, which helped to clear the language of the law, the EU EU Office, which helps to clear the language of the legislation, is concerned that EU EU EU will exacerbate the EU’s “dangerous, informal, informal ways. The leading AII sellers recently opened lobbyists against the EU EU ACT Act, and the commission can now relate to the Trump Office, which is open to the EU movement as Anti -Noviation and Anti-American.

EU deputies receive the third project of the Code of the EU this month, which is compulsory in front of the EU Act and the liabilities that present them as “completely voluntary”. These liabilities include test models to see how they can allow such things such as large-scale discrimination and disinformation spread.

On Tuesday, the President of the European Commission and the President of the President Henna Virkkunen was the first in a letter sent to Virkkunen Financial times However, for the first time, the present and former MPs, the current and former MPs can voluntarily allow the European election, to restrict the freedom of information and violate the EU’s economy and violate the EU’s economy.

“In the current geopolitical situation, it is more important for the increase in the call of the EU and strongly stops strong law and democracy,” he said.

Brando Benifei said the European Parliament’s first signature of the first signature of the European Parliament and the first signature of the European Parliament Fortune Wednesday, the political climate can be connected to the water code. Is the second Trump management antagonist Towards European technological arrangements; Presidential Assistant JD Vance notier In February, the Paris AI would be a “terrible mistake” for European countries in a speech in a speech in the peak of AI.

“I think there is pressure from the United States, but not going in this direction, saying that we can never go to this direction, he will never go in this direction, he said that the European Parliament is currently chaired by the United States

Benifei, O and other former ACT acts, Tuesday, the Practice Code, said he met with AI office experts from the Commission, which prepared an internship codex. Based on this meeting, criminal changes could be returned before the code is finalized.

“I think the issues we raise were considered and therefore there is a place for development,” he said. “We’ll see it in the next weeks.”

Virkkunen, nor Benifei’s pressure did not respond to the description of our pressure. But there is He previously insisted EU technological rules apply to companies from any country sufficient and consistently. Race Commissioner Teresa Ribera also also kept the EU “human rights (or) democracy and values” to place us “

Payments

The main part of the AI ​​Law is here Article 55An important obligations on the provider of general-purpose AI models of “systematic risk” are actual or reasonable effects on significant liabilities, or “public health, security, public safety, fundamental rights or society.” “

ACT says that the calculation capacity used in a model training is more than 10 “If the flops) are measured by” flops), it is likely to have a model risk of a model^Scorpion“This Includes probability Many of the most powerful AI models today, although the European Commission advises scientific advisers to do so, any commission can determine if there is systematically risky model.

According to the law, the providers of such models must evaluate any system risks to “identify and relieve.” This assessment has to work with other words, other words, other words, the model to understand that the model needs to be protected against evil. Then they must assess the European Commission’s AI office and what they found.

Where is this The third version of the Practice Code becomes problematic.

The first version of the code was open to the fact that due to the threat of AI companies, democratic values ​​and election intervention, they need to approach large-scale disinformation or misconceptions when evaluating their models. The second version did not speak specifically about disinformation or misrepresentation, but it was still a systemic risk of “large-scale manipulation for” basic legal or democratic values ​​”,”, for example.

Both first and second versions were clear that model providers should consider the possibility of a large-scale discrimination such as systematic risk.

However, the third version only lists the risks in democratic processes and Basic European Rights As discrimination, discrimination as “for potential consideration in the selection of systemic risks.” The official summary of changes in the third project protects that these are “additional risks that providers can choose to evaluate and soften the future.”

In this week’s letter, MPs negotiating with the commission on the last text of the law insisted that “this is never intentioned.”

“The risks of the main law and democracy are the most risks to assess and soften the most effective AI providers.” “It creates a legitimate uncertainty to comment and narrow the legal text that has agreed on dangerous, non-democratic and legislation.”

This story was first displayed Fortune.com

Source link