Claude Now Marks AI-Generated Text: How Anthropic’s Invisible Watermarks and C2PA Provenance Work

Introduction

Anthropic has begun rolling out machine-readable marking for supported Claude-generated content.

The change has two layers:

1. An embedded watermark in generated text.
2. Digitally signed provenance metadata for supported generated files.

The policy is tied to Anthropic’s commitments under the European Union’s AI Act Article 50(2) Code of Practice on Transparency of AI-Generated Content.

But the implementation is broader than the EU.

Anthropic says that when a Claude model supports marking, the marks apply wherever that model is offered worldwide. The same model-level text marking is intended to appear whether the output comes from Claude’s consumer interface, the API, Claude Code, Claude Cowork, Claude Tag, or supported cloud-provider integrations.

That does not mean every piece of Claude text ever generated can now be detected.

There are several important limits:

• Models released before the marking rollout may not yet carry marks.
• Heavy rewriting, paraphrasing, translation, or mixing with other text can weaken or remove a detectable signal.
• Very short passages may not contain enough signal for reliable detection.
• File metadata can be stripped by screenshots, format conversion, re-saving, or unsupported platforms.
• A detected Claude mark indicates that content may have been processed by Claude; it does not prove Claude was the original author.

Those distinctions are the difference between a useful provenance signal and a universal AI-authorship detector.

Why Anthropic Is Adding Machine-Readable Marks

The immediate regulatory backdrop is the EU AI Act.

Article 50 introduces transparency obligations for certain AI systems and AI-generated content. The European Commission’s transparency Code of Practice is designed to help providers and deployers satisfy those obligations.

The Commission describes the relevant requirements as covering:

• Machine-readable marking of AI-generated or manipulated content.
• Detection of those markings.
• Labelling of deepfakes.
• Labelling of certain AI-generated or manipulated text published to inform the public.

The transparency obligations became applicable on August 2, 2026.

Anthropic signed the Article 50(2) transparency code as a provider of generative AI models and systems.

Its Claude marking system is the company’s implementation path for those commitments.

The key point is that the legal requirement is about transparency and provenance signals, not about making every AI-generated sentence visibly display a warning to the reader.

Machine-readable marking can operate invisibly in the content while still allowing compatible tools to inspect it.

Claude Uses Two Complementary Marking Layers

Anthropic’s design separates text from file provenance.

Using both layers matters because text and files fail differently.

A metadata-only system is easy to lose when a file is converted.

A text-only signal cannot describe the richer history of a generated image or other file.

Anthropic is therefore using a layered approach rather than treating one technique as sufficient for every output type.

Layer One: An Invisible Watermark Embedded in Text

For supported models, Claude-generated text contains an imperceptible watermark embedded directly in the text.

Anthropic says the watermark:

• Is not visible to the reader.
• Does not change the intended meaning.
• Does not reduce readability.
• Travels with the text when it is copied and pasted.
• May persist through some editing.
• Is applied at the model level rather than only in one product interface.

That last point is especially important.

The watermark is not just an HTML tag added by the Claude website.

Anthropic says it is integrated at the model-output level, which is why supported text can remain marked even when the same model is accessed through different Claude products or platforms.

Copying and Pasting Does Not Automatically Remove It

If a user copies a marked Claude paragraph from one interface into:

• An email.
• A document.
• A CMS.
• A note-taking app.
• A blog draft.

the watermark may travel with the text.

That is different from visible labels or ordinary file metadata, which can disappear as soon as content is copied out of the original application.

However, “travels with the text” should not be interpreted as “cannot be removed.”

Anthropic explicitly says that sufficiently heavy transformation can make the signal undetectable.

What Editing Can Do to the Text Watermark

Anthropic lists several cases where Claude-generated text may no longer carry a detectable mark:

• Heavy editing.
• Paraphrasing.
• Translation.
• Mixing with other writing.
• Very short excerpts.

This is a fundamental limitation of text watermarking.

Text is easy to transform while preserving meaning.

For example:

Original Claude draft
→ human rewrites 30% of it
→ translator changes the language
→ editor combines it with human-written paragraphs
→ final document contains only short excerpts

At some point, there may not be enough of the original statistical or embedded structure for the detector to identify a supported Claude mark reliably.

This means the watermark should be treated as a positive provenance signal when present, not a proof of human authorship when absent.

A Claude Mark Does Not Prove Claude Wrote the Original Content

This is one of Anthropic’s most important caveats.

A detected mark can mean the content was processed by Claude.

That is not the same as proving Claude originated all of its ideas or wording.

Consider a user who writes a report manually and asks Claude to:

• Proofread it.
• Translate it.
• Shorten it.
• Improve clarity.
• Convert it into another file format.
• Summarize it.

The resulting output may carry a Claude mark even though the underlying content began with the user.

The correct interpretation is:

Claude mark detected
→ Claude likely processed this content

not:

Claude mark detected
→ Claude independently authored everything in this content

That distinction matters for schools, publishers, employers, compliance teams, and anyone attempting to infer authorship from an automated signal.

Layer Two: Signed Provenance Metadata for Files

When Claude generates a supported file type, Anthropic says it can attach digitally signed provenance metadata.

Examples include:

.svg
.png
.jpg

The metadata follows the Coalition for Content Provenance and Authenticity (C2PA) standard.

C2PA is an open technical standard for recording the provenance of digital content.

Its Content Credentials can contain signed information about a file’s history, such as:

• The tool or service involved.
• How the asset was created or processed.
• Changes in the content’s history.
• Information bound cryptographically to the asset.

Anthropic says a valid signed label can signal that a file was processed by Claude and help detect whether the provenance information has been tampered with.

What C2PA Does—and Does Not—Mean

C2PA is sometimes described too loosely as “an AI watermark.”

That is not quite right.

C2PA is a provenance standard.

It can be used for AI-generated media, but it can also be used for:

• Camera-captured media.
• Edited news photography.
• Audio.
• Video.
• Documents.
• Other digital assets.

Its purpose is to create a verifiable history of the content.

A useful analogy is a signed provenance record attached to a digital asset.

The record can tell a compatible verifier something about where the asset came from and how it was processed.

It does not make the pixels themselves impossible to copy.

File Metadata Can Still Be Lost

Anthropic explicitly warns that provenance metadata is not indestructible.

It may disappear through:

• Format conversion.
• Re-saving.
• Screenshots.
• Copying only the visual content.
• Export through software that does not preserve the metadata.
• Platforms that do not support that marking type.

For example:

Claude creates a PNG with signed provenance metadata
→ user takes a screenshot of the PNG
→ screenshot becomes a new file
→ original metadata may no longer be present

This is why file provenance and embedded watermarks are complementary.

The provenance layer can carry richer information.

The embedded layer can sometimes survive transformations that remove metadata.

No single layer is perfect.

Where Claude Marking Applies

Anthropic says supported Claude models apply marking across the company’s major product surfaces.

The official coverage list includes:

• Claude Platform / API.
• Claude.
• Claude Code.
• Claude Cowork.
• Claude Tag.

The company also says marking is intended to apply worldwide wherever supported Claude models are offered.

This is broader than a rule that activates only when the user is physically located in the European Union.

The implementation is tied to the model’s marking support.

Once a model supports the system, the output is marked across supported Claude surfaces.

Claude API Outputs Are Included

Developers using Claude through the API are not outside the marking system.

Anthropic explicitly includes Claude Platform (API) in the list of covered products.

For supported models:

Application
→ Claude API
→ generated text
→ embedded Claude watermark

The developer does not need to be using the consumer Claude website for the watermark to be present.

This creates an important compliance question for businesses that use Claude behind their own interface.

The downstream product may:

• Display the text directly.
• Rewrite it.
• Summarize it.
• Combine it with database content.
• Convert it into a document.
• Pass it through another model.
• Send it to an end user.

Anthropic recommends that developers independently evaluate their own Article 50 obligations rather than assuming Claude’s upstream marking automatically solves every downstream compliance requirement.

Claude Code Is Also Covered

Anthropic lists Claude Code as a covered surface.

That means text generated by a supported model through Claude Code can carry the same model-level embedded watermark.

This is relevant because coding-agent output can take several forms:

• Explanations.
• Commit messages.
• Documentation.
• Code comments.
• Generated code.
• Markdown files.
• Terminal-facing text.

The official help page describes the watermark as applying to generated text generally.

However, Anthropic has not yet published enough technical detail to support reliable claims about how the detector will behave on every programming language, short code fragment, minified file, or heavily edited codebase.

Developers should therefore avoid treating the watermark as a software-compliance scanner.

Claude Cowork and Claude Tag Are Included

The same policy extends to Claude Cowork and Claude Tag when they use supported models.

This matters because these products can generate work artifacts beyond a simple chat response.

A user may ask Claude to:

• Draft a report.
• Create a presentation.
• Produce a visual asset.
• Generate a document.
• Prepare structured work output.

Text marking and file provenance can therefore appear in different parts of the same workflow.

A document may contain watermarked text while an exported supported image carries signed provenance metadata.

What Happens on AWS, Google Cloud, and Microsoft Foundry

Anthropic also addresses third-party cloud platforms.

For supported Claude models accessed through:

• Amazon Web Services.
• Google Cloud.
• Microsoft Foundry.

the embedded text watermark is intended to apply.

The file-provenance layer is more conditional.

Anthropic says signed provenance metadata may not be supported on every cloud platform because the available file-processing features differ.

The practical distinction is:

This is another reason not to describe the entire system as one universal watermark.

The text and file layers have different deployment requirements.

Which Claude Models Are Marked?

Anthropic’s current rule is based on model launch date.

The official documentation says:

Claude models launched in the EU on or after August 2, 2026 support machine-readable marking at launch.

Models released before that date are not automatically covered by the same statement.

Anthropic says it is working to add marking support to earlier models.

The company has not published one simple permanent table stating that every historical Claude model now carries both marking layers.

Therefore, a reliable implementation should check current model documentation rather than infer support only from the word “Claude.”

A product using an older model should not assume:

Claude model
= definitely marked

The correct question is:

Does this specific Claude model currently support marking?

The August 2 Date Is Not a Claim That Every Old Output Changed Retroactively

Watermarking applies when supported models generate or process content.

It does not retroactively alter text that Claude produced months earlier.

A document created before the rollout does not suddenly receive a hidden signal because the model is later updated.

Likewise, if an old model did not support marking at the time it generated a passage, the absence of a mark says very little about whether Claude was involved.

That is one reason the detector cannot function as a complete historical AI-authorship database.

Anthropic Is Building Detection Tools

A watermark has limited practical value if nobody can inspect it.

Anthropic says it plans to support detection by:

• Claude users.
• Other third parties.

The company says future technical documentation will explain the detection mechanisms in more detail.

As of the publication date of this article, that technical guidance is still forthcoming.

This means there is an important difference between:

Anthropic has deployed marking support

and:

Anyone can already run a fully documented public detector for every Claude mark

The first statement is supported.

The second is too strong.

What a Future Claude Detector Could Tell You

Anthropic describes detection as checking whether a piece of text or a file carries a supported Claude mark.

If a supported mark is detected, the safe interpretation is:

This content may have been processed by Claude.

That wording matters.

A detector should not automatically claim:

• Claude wrote every sentence.
• No human contributed.
• The content is false.
• The content is plagiarism.
• The user violated a policy.
• The entire document was AI-generated.

Provenance is evidence about processing history.

It is not a moral or quality judgment.

Why “No Watermark Detected” Is Weak Evidence

Anthropic is equally clear about the reverse case.

Failure to detect a mark does not prove that content is human-written.

A Claude-generated passage may lack a detectable signal because:

1. It came from a model that did not yet support marking.
2. It was heavily edited.
3. It was paraphrased.
4. It was translated.
5. It was mixed with other writing.
6. It is too short.
7. The relevant file metadata was stripped.
8. The platform or file type did not support the marking method.

This makes the detector asymmetric.

A positive result can provide useful evidence.

A negative result cannot reliably exclude AI involvement.

Why This Matters for Schools and Employers

AI-detection systems are often used in high-stakes contexts.

Examples include:

• Academic misconduct reviews.
• Hiring assessments.
• Journalism.
• Employee performance reviews.
• Publishing.
• Legal document workflows.

The Claude watermark can improve provenance, but it should not be turned into an automatic guilt detector.

A responsible process should distinguish:

Machine-readable Claude signal

from:

Evidence that a person violated a rule

If a university allows AI-assisted proofreading but prohibits fully AI-written assignments, a detected mark alone cannot determine which activity occurred.

The text may have originated with the student and only passed through Claude for editing.

Human policy and contextual evidence are still required.

Why This Matters for Publishers and Newsrooms

Publishers may find machine-readable provenance useful for internal workflow tracking.

For example, a newsroom could distinguish:

• Human-written draft.
• Claude-assisted editing.
• Claude translation.
• AI-generated illustration.
• Human-edited final publication.

But provenance systems work best when the workflow preserves the data.

If the content is repeatedly copied between tools that strip metadata or heavily rewritten, the chain becomes incomplete.

For sensitive publishing workflows, organizations may need to preserve:

• Original files.
• Content Credentials.
• Editing history.
• AI-assistance logs.
• Human approvals.

Watermarks are most useful as one signal in a broader provenance system.

The System Is Not Designed to Make AI Text “Undeletable”

The source article’s headline suggests that AI-generated content may become impossible to hide.

That overstates what Anthropic says.

Anthropic explicitly documents several ways marks can become undetectable.

Heavy rewriting can weaken the text signal.

Screenshots and format conversions can remove file metadata.

Unsupported surfaces may not preserve every marking type.

The more accurate conclusion is:

Claude-generated content is becoming easier to identify when its provenance signals survive, but the system does not make AI involvement permanently or universally detectable.

This is consistent with the broader technical reality of provenance systems.

Robustness is a spectrum, not an absolute guarantee.

How Claude’s Approach Compares with C2PA and SynthID

Anthropic’s approach combines two ideas that are already visible elsewhere in the AI industry.

C2PA / Content Credentials

C2PA provides cryptographically signed provenance information for digital assets.

The standard is supported across a growing ecosystem of:

• AI companies.
• Camera manufacturers.
• software vendors.
• media organizations.
• content platforms.

The public Content Credentials Verify tool can inspect supported files for compatible credentials.

Google SynthID

Google DeepMind’s SynthID embeds imperceptible watermarks directly into AI-generated content.

Google says SynthID can be used across:

• Images.
• Audio.
• Text.
• Video.

The concept is closer to Anthropic’s embedded watermark layer than to C2PA metadata.

OpenAI Provenance Signals

OpenAI currently documents provenance signals for supported generated media, including C2PA metadata and SynthID in supported image and audio workflows.

OpenAI’s public verification tool is designed for supported media files, not as a universal detector for arbitrary AI-generated text.

The broader industry direction is therefore toward layered provenance:

Embedded signal
+
signed metadata
+
verification tool
+
platform labels

Anthropic’s addition of a text watermark brings Claude more directly into that pattern.

Text Watermarking Is Harder Than Image Metadata

Text creates a particularly difficult provenance problem.

An image can retain the same visual object while carrying hidden information in pixels or metadata.

Text is routinely transformed at the semantic level.

A user can:

• Change synonyms.
• Reorder sentences.
• Translate the text.
• Summarize it.
• Expand it.
• Combine multiple drafts.
• Ask another model to rewrite it.

The result may preserve the same meaning while replacing most of the original token sequence.

A robust text watermark therefore has to balance competing goals:

1. It must remain imperceptible.
2. It must preserve language quality.
3. It should survive ordinary editing.
4. It should be detectable.
5. It should resist false positives.
6. It should not be easily forged.

Anthropic has not yet published enough implementation detail to evaluate all of those tradeoffs independently.

Why Anthropic Is Not Publishing Every Technical Detail Yet

Watermark systems face an adversarial problem.

If an attacker knows exactly how a watermark is constructed, they may attempt to:

• Remove it.
• Forge it.
• Transfer it to human-written content.
• Design transformations that defeat detection.
• Trigger false positives.

At the same time, a system that is completely opaque is difficult for researchers and regulators to evaluate.

Anthropic’s current position is an intermediate one.

It publicly documents:

• The existence of the marking.
• The two-layer architecture.
• Product coverage.
• Major limitations.
• Plans for third-party detection.

Detailed detection guidance is still pending.

That gives the company room to publish a more complete technical framework later.

EU AI Act Article 50 Is Broader Than Claude

Anthropic’s marking system should be understood as one provider’s implementation of a wider European transparency regime.

The European Commission says Article 50 addresses transparency for providers and deployers of generative AI systems.

The Code of Practice is voluntary as a compliance mechanism, but the underlying Article 50 legal obligations are not merely voluntary.

The code includes separate sections for:

• Providers: marking and detection.
• Deployers: labelling deepfakes and certain AI-generated or manipulated text.

The practical obligations can therefore differ depending on whether an organization:

builds the generative AI system

or:

uses the generative AI system to publish content

Developers should not assume that using a compliant upstream model automatically satisfies every obligation imposed on the downstream deployer.

The Transparency Code and the GPAI Code Are Different

There are multiple EU AI codes of practice, which can create confusion.

The Code of Practice on Transparency of AI-Generated Content relates to Article 50 marking and labelling obligations.

The earlier General-Purpose AI Code of Practice addresses obligations for general-purpose AI model providers, including areas such as:

• Transparency documentation.
• Copyright.
• Safety and security for systemic-risk models.

Anthropic is a signatory to the GPAI Code as well.

The Claude marking article specifically references the Article 50(2) transparency code.

Those two compliance frameworks should not be treated as the same document.

What Developers Using Claude Should Do

Anthropic’s advice is straightforward: assess your own obligations independently.

A practical implementation checklist looks like this.

1. Identify the Claude Model

Record the exact model used by your application.

Do not rely on a generic “Claude” label.

1. Confirm Marking Support

Check whether the model currently supports machine-readable marking.

Older model behavior may differ.

1. Map Your Output Pipeline

Document what happens after Claude returns content.

For example:

Claude API
→ application post-processing
→ database
→ another model
→ template engine
→ PDF export
→ customer

Every transformation can affect provenance.

1. Separate Text and File Provenance

Do not assume that C2PA metadata and text watermarking survive the same operations.

Test them separately.

1. Preserve Original Artifacts When Needed

For regulated or high-trust workflows, retain:

• Original model output.
• Original generated file.
• Signed metadata.
• Transformation history.
• Human edits.
• Final publication.

1. Do Not Treat Detection as Authorship Proof

A positive mark can show Claude processing.

It may not prove original authorship.

1. Test Short and Edited Content

Your use case may involve:

• Headlines.
• Code snippets.
• Summaries.
• Translations.
• Heavily edited reports.

These are exactly the cases where detection may be less reliable.

1. Review Article 50 Obligations

If your product is offered in or affects the EU market, assess the law and current Commission guidance for your specific role.

1. Watch for Anthropic’s Detection Documentation

Anthropic says more technical guidance is coming.

Do not build a permanent compliance architecture around assumptions about an unpublished detector.

A Practical Provenance Workflow

For a company that uses Claude to create public content, a defensible workflow could look like this:

Human or source data
↓
Claude supported model
↓
Original marked output saved
↓
Human editing
↓
Final publication file
↓
Provenance metadata preserved where supported
↓
Disclosure / label added where legally or editorially required
↓
Audit record retained

This approach does not depend on the watermark surviving every transformation.

It preserves an internal evidence chain even when the public artifact changes.

What Content Creators Should Expect

If you use a supported Claude model to draft:

• Emails.
• Reports.
• Blog posts.
• Documentation.
• Marketing copy.
• Articles.

the generated text can contain a machine-readable watermark.

Copying the text into another application does not necessarily remove it.

That does not mean your visible document will display “Written by Claude.”

The mark is intended to be imperceptible.

It also does not mean every future reader can immediately scan it today.

Anthropic’s third-party detection guidance is still being developed.

What Happens After Human Editing?

There is no single percentage of editing after which the watermark disappears.

Anthropic does not publish a universal threshold such as:

20% rewrite = detectable
50% rewrite = undetectable

The company only says the mark may survive some edits and can become undetectable after heavy editing, paraphrasing, translation, or mixing.

The outcome can depend on:

• Passage length.
• Type of edits.
• Amount of unchanged text.
• Detector design.
• Language.
• Transformation method.

Any tool claiming a precise universal edit threshold should therefore be treated cautiously unless Anthropic publishes evidence supporting it.

What Happens to Generated Images?

For supported generated file types such as PNG and JPG, Claude can attach signed provenance metadata following C2PA.

If the metadata remains intact, compatible verification tools can inspect it.

If the image is:

• Screenshotted.
• Converted.
• Re-exported through unsupported software.
• Uploaded to a service that strips metadata.

the provenance information may disappear.

A missing credential is not proof that an image was human-created.

What Happens to SVG Files?

Anthropic specifically includes SVG among its examples of supported file types for signed provenance metadata.

SVG is useful because Claude can generate vector graphics and diagrams as structured files.

Like other file formats, provenance survives only when the processing chain preserves the metadata.

A user who copies only the SVG markup into a new file may not preserve the original credential.

Can Someone Forge a Claude Watermark?

Anthropic’s current public documentation does not provide enough information to make a reliable claim about the difficulty of forging its embedded text mark.

This is an important open security question.

Any provenance system has to consider both:

removal attacks

and:

forgery attacks

A malicious actor could benefit from making human content appear AI-generated just as they could benefit from hiding AI origin.

That is one reason high-stakes decisions should not rely on one automated signal.

Will Watermarks Stop AI Misinformation?

No.

Machine-readable marking can improve transparency, but it cannot solve misinformation on its own.

Bad actors can use:

• Models without marking.
• Older models.
• Open-weight models with modified inference code.
• Human rewriting.
• Screenshots.
• Metadata-stripping tools.
• Entirely human-written false claims.

Provenance helps answer:

What can we learn about how this content was created or processed?

It does not answer:

Is this content true?

Fact-checking and provenance are complementary tasks.

The More Useful Future Is Provenance, Not “AI Detection”

The phrase “AI detector” suggests a binary test:

AI
or
human

Real content creation is becoming more mixed.

A document may involve:

• Human research.
• Claude drafting.
• Human editing.
• AI translation.
• Human approval.
• AI formatting.
• Human publication.

In that world, the more useful question is not:

“Was AI used?”

It is:

“What happened to this content, which tools were involved, and which parts of the chain can be verified?”

That is the problem provenance standards are designed to address.

常见问题

Does Claude now watermark all generated text?

Supported Claude models can embed machine-readable watermarks in generated text. Anthropic says models launched in the EU on or after August 2, 2026 support marking at launch, while marking support for earlier models is still being added.

Can I see the Claude watermark in the text?

No. Anthropic describes the watermark as imperceptible and says it does not change the intended meaning, quality, or readability of the response. It is designed for machine-readable detection rather than visible labelling.

Does copying Claude text into Word or email remove the watermark?

Not necessarily. Anthropic says the embedded signal travels with copied and pasted text and may survive some editing. Heavy rewriting, paraphrasing, translation, mixing, or shortening can make the mark undetectable.

Can a Claude watermark prove that Claude wrote the entire document?

No. A detected mark can indicate that the content was processed by Claude, but Claude may have only proofread, translated, summarized, or reformatted human-written material. Anthropic explicitly warns that the mark is not full proof of original authorship.

Does Claude add C2PA metadata to generated images?

Anthropic says supported generated file types such as SVG, PNG, and JPG can receive digitally signed provenance metadata following the C2PA standard. Support can depend on the product, platform, feature, and file type.

Are Claude watermarks active through AWS, Google Cloud, and Microsoft Foundry?

Anthropic says embedded text watermarks apply when supported Claude models are accessed through AWS, Google Cloud, or Microsoft Foundry. Signed provenance metadata may not be supported on every cloud platform because file-processing capabilities differ.

Is there a public Claude watermark detector?

Anthropic says it is building detection support for users and third parties and will publish more technical guidance. As of August 11, 2026, the company’s official documentation does not yet provide a complete public detection specification for the embedded text watermark.

Does the absence of a Claude watermark prove that text was written by a human?

No. A mark may be absent or undetectable because the model predates marking support, the passage is too short, the text was heavily edited or translated, or the relevant file metadata was stripped. Absence of a signal is not reliable proof of human authorship.

相关工具

• Claude: Anthropic’s official AI assistant and one of the product surfaces covered by machine-readable marking for supported models.
• Claude Platform: Anthropic’s developer platform for building applications with Claude models, including supported marked-model outputs.
• Claude Code: Anthropic’s coding agent, whose supported-model text outputs fall under the same model-level marking policy.
• Content Credentials Verify: An official verification tool for inspecting supported Content Credentials and provenance information in digital assets.
• C2PA: The standards organization behind the open Content Credentials provenance specification used by Claude’s signed file metadata.
• Google SynthID: Google DeepMind’s embedded watermarking technology for AI-generated text, images, audio, and video.
• EU Transparency Code of Practice: European Commission guidance supporting compliance with AI Act Article 50 marking and labelling obligations.

Related Links

• Anthropic: How Claude Marks AI-Generated Content: The primary official source for Claude’s text watermarks, C2PA provenance, product coverage, cloud behavior, detection plans, and limitations.
• EU Code of Practice on Transparency of AI-Generated Content: European Commission page explaining the provider and deployer transparency framework under Article 50.
• EU Artificial Intelligence Act: The official legal text establishing the EU’s AI transparency obligations and broader regulatory framework.
• C2PA Technical Specification: The open technical specification for cryptographically bound provenance records and Content Credentials.
• C2PA Content Credentials Explainer: A more accessible explanation of how provenance records are attached, signed, and verified.
• Content Credentials Verify: A public tool for examining compatible provenance credentials in supported digital content.
• Google DeepMind SynthID: Official overview of Google’s imperceptible watermarking system across multiple generative media types.

Summary

Anthropic is introducing machine-readable provenance for supported Claude outputs through two complementary mechanisms: an imperceptible watermark embedded in generated text and signed C2PA-based provenance metadata attached to supported files.

The rollout is tied to the EU AI Act’s Article 50 transparency framework, but Anthropic says supported models apply marking worldwide across Claude, the API, Claude Code, Claude Cowork, Claude Tag, and supported cloud-provider access paths. New models launched in the EU on or after August 2, 2026 support marking at launch, while earlier models are still being updated.

The system is useful, but it is not a perfect AI-authorship detector. Text marks can become undetectable after heavy editing or translation, file metadata can be stripped, and a detected signal only shows that Claude may have processed the content—not that Claude originated every part of it.

The practical shift is from guessing whether something “looks AI-written” toward preserving verifiable evidence about how digital content was created and processed.