On August 2, 2026, two artificial-intelligence transparency laws take effect on the same day, on two continents. In Europe, the EU AI Act transparency rules bind. In the United States, California’s AI Transparency Act binds. That shared date is not a coincidence, and the timing is worth getting right, because the headlines have been getting it wrong.
The correction first: no, the EU did not delay this
You may have read that the EU delayed its AI Act. It delayed one part. Late in 2025 the bloc’s “Digital Omnibus” pushed back the high-risk obligations, the Annex III rules, from August 2026 to December 2027. The transparency rules, Article 50, were not moved. They still bind on August 2, 2026, and the fining power turns on the same day. The penalty tier for breaking them runs up to 15 million euros or 3 percent of worldwide annual turnover, whichever is higher.
So the part of the law that touches the most people, transparency, is exactly the part that stayed on schedule. Anyone waiting on the “delay” is planning around a date that does not apply to them.
What Article 50 actually asks for
In plain terms: if you provide an AI system that generates synthetic images, audio, video, or text, the outputs have to be marked in a machine-readable way that says a machine made them. And if you deploy AI to create or alter a deepfake, or to generate text published to inform the public on a matter of public interest, you have to disclose that it is AI. It is a labeling and disclosure duty, aimed at the people putting AI content in front of an audience.
California, the same day, on purpose
California’s AI Transparency Act (SB 942) was originally set for January 1, 2026. A 2025 amendment (AB 853, signed October 13, 2025) moved it to August 2, 2026, explicitly to line up with the EU’s Article 50 timeline. It aims at the large providers, those with more than one million monthly users in California, which is effectively every major AI lab. It requires three things: a permanent, machine-readable watermark on generated images, audio, and video; a free public tool that lets anyone check whether a piece of content came from that provider; and an optional visible “AI-generated” label users can attach. Text-only outputs are excluded from the watermark rule.
What it means for you
Two practical shifts. First, if you publish AI-made content to European audiences, you now have disclosure duties, and there is a real penalty attached. Label it. Keep the provenance data intact instead of stripping it. Second, the big AI tools you already use will start stamping their outputs and handing you free detection tools, which means “is this actually AI?” becomes a question you can check instead of guess. The honest posture underneath both laws is the same one worth holding regardless of any statute: disclose what a machine made, and verify before you trust.
We are logging this pair, dated, on the record, and we will note who actually complies. If you want to pressure-test any claim about what these laws do or do not require, run it through Grade This Claim, or browse the rest of the free tools.
Sources: EU AI Act Article 50 and Article 99 (artificialintelligenceact.eu; compliance analyses via Gibson Dunn and Sidley); California SB 942 as amended by AB 853 (California Legislative Information; Troutman Pepper). Dates and figures verified against primary and authoritative secondary sources, July 2026. Educational only, not legal advice; confirm against the statutes before acting. Real numbers. No hype. Receipts.
Leave a Reply