Brussels has released a practical guide to assist companies in preparing for the upcoming enforcement of the AI Act. The document outlines detailed requirements for general-purpose AI models across three key areas: transparency, copyright, and safety.
Developers are tasked with documenting data sources used for training, providing audit interfaces, and implementing filters to protect copyrighted content. Safety requirements mandate the conduct of red-teaming exercises and risk assessments to identify potential vulnerabilities.
Also read:
- CZ Proposes a New Era for Decentralized Exchanges: Introducing Dark Pool Perpetual DEX
- Tesla's Optimus Robot Masters Household Tasks
- AI Trends Report by BOND: Mary Meeker’s Comeback Highlights Unprecedented AI Revolution
These regulations will become mandatory on August 2, 2025, with the publication signaling the regulator’s refusal to delay timelines despite business requests for extensions. Non-compliance could result in fines of up to €35 million or 7% of a company’s annual turnover.
