EU countries and lawmakers have reached a provisional agreement on the AI Act, which notably softens the initial framework set to regulate artificial intelligence in the region. The deal postpones the implementation of rules governing high-risk AI systems—such as biometric applications and those used in law enforcement—until December 2025. Additionally, the agreement introduces a ban on unauthorized sexually explicit AI-generated images and mandates watermarking for AI outputs, both set to take effect on December 2, 2026. Notably, machinery has been exempted from this regulation, allowing it to continue under existing sector-specific regulations.

EU countries: The 27 member states of the European Union that collectively participate in legislative and regulatory decisions affecting the bloc. In this news, EU countries negotiated with European Parliament lawmakers to reach a provisional agreement on modified artificial intelligence regulations.
European Parliament lawmakers: Elected representatives of the European Parliament who draft and approve EU legislation on behalf of European citizens. They agreed with EU governments on Thursday to revise the AI Act’s implementation timeline and scope, including delays to high-risk AI system enforcement.

`json
{
“Regulatory Scope”: “Ensure any exclusions of machinery from the AI Act are supported by current regulations, as it could be considered misinformation otherwise.”,
“High-Risk AI Coverage”: “High-risk AI systems include biometric systems, critical infrastructure applications, and law enforcement tools, with their regulation under the updated agreement.”
}
`