Anthropic Watermarks Text. The Grok Suit Is About a Photo.

The AI governance that arrived on a schedule this week is the part that was easiest to specify. Anthropic published a technical FAQ explaining that its new text watermarking exists to satisfy the EU AI Act's Code of Practice, whose August deadline carries penalties up to €15M or 3% of global turnover. In the same news cycle, the Washington Post and TechCrunch reported a Wyoming lawsuit alleging a man used Grok to turn a stepdaughter's childhood photograph into sexually explicit imagery — a generation-side harm that no labeling regime addresses, regardless of modality. And the most substantive safety assessment published this week, Anthropic's second Responsible Scaling Policy Risk Report, is self-published on a cadence the company set for itself, because no external body publ ...