Technology · 18 August 2026

Claude AI's new text watermark has a simple workaround

Anthropic, the tech firm behind the assistant, just began hiding invisible signals inside its global responses to satisfy European transparency rules. But when City AM asked the system about the change, it admitted people could simply translate or rewrite the text to mask its origin. The new feature tweaks how the bot selects vocabulary, using a method borrowed from Google DeepMind instead of obvious tags.

By BriefTea Editorial · 18 August 2026 at 16:18

Reported by City AM · How we write briefs

Get the app

Understand this story

What are AI watermarks? Can AI watermarks be removed? Who is Anthropic?

More briefs

GeneralSouthern England declared in severe drought across six areas TechnologyTesla's new Roadster: 60 mph in under a second, thanks to SpaceX WorldTrump administration places new sanctions on 26 targets connected to Iran PoliticsChancellor John Healey plans to raise taxes on banks in the October budget BusinessLondon's top shares reach a seven-week high as investors buy safer British businesses GeneralHarry and Meghan visit Nottingham in their first joint public outing since returning Browse all stories →