Recently, EU member states and Members of the European Parliament reached a preliminary agreement on a "watered-down" AI Act, marking an important development in the EU's digital regulation. This act is part of the European Commission's "Digital Package."
After long late-night negotiations, the parties ultimately decided to delay the compliance time for high-risk AI systems until December 2027, meaning the original effective date of August 2, 2023, has been postponed. These high-risk systems include biometric technologies, critical infrastructure, education, employment, law enforcement, and border management.
According to the agreement, machinery will no longer be subject to this act, as these devices should follow existing industry safety regulations. This decision has been criticized by opponents as a compromise to corporate pressure. At the same time, the mandatory watermarking requirement, which will take effect on December 2 this year, aims to enhance the transparency and credibility of AI-generated content.
Additionally, the EU has reached a consensus to ban the use of AI technology to create unauthorized pornographic deepfake content, including images, videos, and audio. Relevant companies need to complete system compliance adjustments by the end of the year to comply with new legal requirements. This ban also specifically emphasizes the protection of children, clearly including the production of fake images related to child sexual abuse in the regulatory scope.
This agreement is seen as an important measure by the EU to improve legal certainty and reduce administrative costs for businesses, helping to ensure the uniform implementation of relevant regulations across member states. Marie-Lena Raouf, Deputy Minister of European Affairs of Cyprus, stated that this agreement will strongly support local businesses and enhance the EU's competitiveness in the digital economy.
Although this act still needs formal approval from the European Parliament plenary session and the governments of EU countries, this process is typically considered a routine procedure and is expected not to face major obstacles.
Key Points:
✅ The EU has decided to delay the compliance time for high-risk AI systems until December 2027.
🚫 Prohibits the use of AI to create unauthorized pornographic deepfakes, protecting children.
📅 Mandatory watermarking requirements will take effect this December, increasing the transparency of generated content.
