EU explains how to do AI without breaking the law
Briefly

The European Commission has issued the General-Purpose AI Code of Practice to aid companies in meeting compliance with the forthcoming AI Act. The Code is voluntarily signable by companies and is divided into three parts. It addresses responsibilities related to transparency and copyright as well as a comprehensive section on safety and security, applicable to certain advanced AI models. Regulations include avoiding copyrighted outputs, adherence to robots.txt by AI scraping bots, and conducting risk assessments. Companies have additional requirements for documentation of model details, which must be retained for ten years.
The European Commission announced the publication of the General-Purpose AI Code of Practice, aimed at helping companies comply with the AI Act. Compliance is voluntary.
The Code includes responsibilities for AI companies regarding transparency, copyright adherence, and safety/security, primarily targeting advanced model providers in critical sectors.
Key regulations include preventing output of copyrighted content, requiring adherence to robots.txt for AI scraping bots, and enforcing risk assessments, along with documentation requirements.
Companies must document all model intricacies, maintain records for a decade per version, and search engines are advised against downranking non-compliant pages.
Read at Theregister
[
|
]