The lawmakers also pointed out that the draft code’s current language could allow AI providers to focus on only the risks that are easiest to mitigate, rather than the most serious risks that could have a major impact on society as a whole. They warned that this could lead to AI providers prioritizing their own interests over the public good.
“If this approach is maintained, it will undermine the very purpose of the obligations set out in Article 55 of the AI Act,” they wrote.
Benifei echoed these concerns in his conversation with Fortune, saying that the AI Office’s current approach could allow AI companies to “cherry-pick” which risks they assess and mitigate, rather than taking a comprehensive approach to protecting fundamental rights and democracy.
“We are talking about models that are very powerful and have a risk of causing harm at a very large scale, so the idea that you could pick and choose which risks to assess and mitigate—it’s not coherent with the goal of the AI Act,” he said.
The lawmakers also called on Virkkunen to ensure that the final code of practice aligns with the intent of the AI Act and reflects the agreement that was reached during negotiations. They emphasized the importance of protecting fundamental rights and democracy in the face of powerful AI models that could potentially be misused.
“We urge you to ensure that the final version of the code of practice reflects the obligations laid down in the AI Act, so that the most impactful AI models are subject to the necessary safeguards to protect fundamental rights and democracy,” they wrote.
As the European Commission works to finalize the code of practice in the coming weeks, it remains to be seen whether the concerns raised by the lawmakers will be addressed. The outcome could have significant implications for how AI providers evaluate and mitigate the risks associated with their models, and ultimately, for how AI is regulated in the EU.
Es ist gefährlich, undemokratisch und schafft rechtliche Unsicherheit, einen Rechtstext, auf den sich die Mitgesetzgeber geeinigt haben, vollständig neu zu interpretieren und einzuschränken, durch einen Verhaltenskodex.
Diese Geschichte wurde ursprünglich auf Fortune.com vorgestellt