Bottom line: AI providers must now label AI-generated content and inform users about chatbot interactions; penalties for non-compliance reach up to €35 million or 7 percent of global annual turnover.
With the EU AI Act, enhanced labelling requirements for AI-generated content such as deepfakes, videos and chatbot interactions now apply. The European Commission has published concrete specifications for marking and provides for substantial penalties if providers breach the rules.
Companies creating artificial intelligence must now systematically label when their models generate visual, audio or textual content. For AI videos, watermarks are envisioned at the beginning, during the course or at regular intervals. For manipulated images, in addition to permanent marking, interactive elements such as hover information should alert users to the artificial origin. For voice content, a spoken notice is required at the beginning; for text, AI labels above the text or at the heading. In addition, machine-readable fingerprints are to be embedded in the files to enable automatic detection.
The ruleset distinguishes between new and already established models. All newly launched AI systems are subject to the requirements immediately. For widely distributed models such as ChatGPT, Gemini or Claude, a transitional period applies until December 2026 at the latest. Private individuals are exempt from the requirements provided they do not use AI for professional purposes.
For Chief Data Officers and compliance officers, it is relevant that the European AI Office in Brussels will in future examine whether large language models comply with security standards and cannot conduct autonomous cyberattacks. The authorities demand appropriate security measures and immediate reporting of security incidents. Concrete background: OpenAI and Anthropic had both confirmed in test scenarios that their systems had unintentionally attacked external companies.
Breaches of the law are subject to substantial sanctions. For prohibited AI practices, penalties of up to €35 million or 7 percent of global annual turnover may be imposed, whichever is higher. For violations of transparency obligations or inadequate security measures, penalties of up to €15 million or 3 percent of global annual turnover are possible.
Source: www.it-daily.net · Published 2 August 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.