Search+ET Online | 12 Aug 2026, 12:33 PM ISTIf you wrote it with Claude, it may now carry an invisible, machine-readable watermark inside the text.Anthropic has introduced a text watermarking system for Claude-generated content that travels with the words when they are copied and pasted. The marking applies across Claude’s app, API, Claude Code, Claude Cowork and supported cloud platforms. Files can also carry C2PA provenance metadata.The move is linked to EU AI Act transparency requirements, but the marking applies globally, meaning Indian users are already affected.There is an important India-specific distinction, however. India’s IT Amendment Rules require labelling and provenance metadata for synthetically generated information, while MeitY’s clarification has focused on multimodal content such as images, audio and video rather than ordinary text.And the detection tool itself has not shipped yet. Heavy editing, paraphrasing or translation can weaken or destroy the signal. Even if a text is flagged, that does not prove Claude originally wrote it. Using Claude to proofread or summarise your own draft can leave the same trace.For students, freelancers and newsrooms, the takeaway is simple: *treat an AI-detection flag as a signal, not proof of authorship.* Show more Show lessRead more onRelated Videos00:21“Why did PM Modi inaugurate it?” Priyanka Gandhi's sharp attack on PM over Ram Mandir theft case“At least wear proper clothes…” CM Rekha Gupta rebukes MLA in Assembly00:39“I was at greater risk…” Trump opens up on covert Air Force One switchIndia-US Tariff row: ‘PM Modi-Trump will work it out’00:21“I got firebombed…” Trump adviser Peter Navarro on backlash over India remarks01:1301:3200:3901:0500:3301:0000:3701:0800:3102:2501:3701:0401:1600:2400:5001:2000:5601:3101:1301:0401:0101:0201:4501:3701:4300:5301:1000:4801:4301:4400:4101:03
Claude watermark explained — What Indian users need to know
If you wrote it with Claude, it may now carry an invisible, machine-readable watermark inside the text.Anthropic has introduced a text watermarking system for Claude-generated content that travels with the words when they are copied and pasted. The marking applies across Claude’s app, API, Claude Code, Claude Cowork and supported cloud platforms. Files can also carry C2PA provenance metadata.The move is linked to EU AI Act transparency requirements, but the marking applies globally, meaning Indian users are already affected.There is an important India-specific distinction, however. India’s IT Amendment Rules require labelling and provenance metadata for synthetically generated information, while MeitY’s clarification has focused on multimodal content such as images, audio and video rather than ordinary text.And the detection tool itself has not shipped yet. Heavy editing, paraphrasing or translation can weaken or destroy the signal. Even if a text is flagged, that does not prove Claude originally wrote it. Using Claude to proofread or summarise your own draft can leave the same trace.For students, freelancers and newsrooms, the takeaway is simple: *treat an AI-detection flag as a signal, not proof of authorship.*
Anthropic deployed machine-readable watermarks for Claude text across platforms (app, API, Code, Cowork) to comply with EU AI Act, applied globally. IT teams must audit AI-generated assets; detection is unreleased and editing weakens the watermark.
