The European Commission published detailed guidance on 16 August 2024 clarifying the obligations of general-purpose AI model providers under the AI Act, which enters into application in stages beginning in early 2025. The guidance addresses transparency requirements, including disclosure of training data and compliance with copyright law, and systemic-risk evaluation for the most capable models. Providers must document risk-mitigation measures, implement internal governance and cooperate with national authorities, with particular attention to dual-use capabilities that could facilitate the generation of illegal content, including child sexual abuse material.
When a general-purpose model can be weaponised to produce synthetic child sexual abuse material, detection must operate downstream at the point of distribution. Guardii's CSAM detection module, which identifies both photographic and AI-generated abuse imagery in real time within direct messages on Instagram, Snapchat, Discord and Roblox, blocks such material before it reaches a child and enables immediate escalation to law enforcement. As the AI Act places upstream obligations on model developers, a Meta Business Partner platform with live pattern recognition closes the loop at the point of contact, intercepting prohibited outputs in the channels where harm occurs and surfacing actionable intelligence to the right authority without delay.
