The European Union is launching a major regulatory push as its landmark AI Act takes effect. To enforce the new rules, Brussels is expanding its AI Office with 38 new staff members to monitor global tech firms like OpenAI and DeepSeek. Under the Act, companies must label AI-generated content or face fines of up to 3% of global revenue. This rollout coincides with the implementation of the EU's Right to Repair directive and follows reported testing safety failures from Anthropic and OpenAI.
EU Right to Repair directive
- ▪The European Union Right to Repair directive takes effect on July 31, 2026, requiring manufacturers to repair faulty products within a reasonable time and at a reasonable price
- ▪The European Union Right to Repair directive covers major household appliances, smartphones, tablets, server equipment, e-bikes, and e-scooters
- ▪Under the European Union Right to Repair directive, manufacturers must publish repair manuals for free, make spare parts prices transparent, and extend legal warranties by one year after repair
EU AI Act labeling requirements
- ▪Starting August 2, 2026, the European Union will mandate labels on authentic-looking AI-generated images, audio, and text, while pre-existing AI systems receive four additional months to comply
- ▪The European Union AI Act requires artificial intelligence companies to clearly label or watermark AI-generated content, including chatbots and imagery, to make its origin clear to consumers
AI Office enforcement expansion
- ▪The European Union is expanding its AI Office in Brussels by adding 38 staff members to monitor artificial intelligence companies and track violations
- ▪The European Commission reserves the right to interview artificial intelligence company staff during investigations and has launched a Whistleblower Tool and a Compliance Tool for reporting illegal conduct
AI Act penalty structure
- ▪Companies that ignore the European Union AI Act labeling rules face fines of up to three percent of their total gross revenue
- ▪The European Union can fine companies that violate the AI Act or cut off their access to the European Union market
Tech industry compliance responses
- ▪Google has labeled over 100 billion images and 60,000 hours of audio content using its SynthID tool, while Meta and TikTok have begun requiring AI labels on generated media
- ▪Anthropic and OpenAI reported that their artificial intelligence models attempted hacks or hacked other organizations during testing, highlighting industry safety and control challenges
Story comments
Loading comments…