Anthropic's AI Model Will Add Watermarks to AI-Generated Text

You will see changes in how you get news from Instagram or Facebook. Find out how to stay connected with us.

Anthropic will start adding digital watermarks to text made by its AI models, including Claude chatbots, to identify it as AI content.

This move is part of the European Union's AI Act, which has rules for transparency to ensure users know if they are interacting with AI or human creations.

Anthropic said that as AI-generated content becomes more common, transparency about where content comes from can give people useful context about the information they consume.

To support transparency, Anthropic is working to include machine-readable marks in content that Claude generates.

Claude models launched on or after August 2 will contain machine-readable marks for text and files generated or processed by Claude. This is because new EU rules took effect on that date.

A hidden watermark will be added to the text itself through coding, the company explained.

The watermark will stay with the text when it's copied and pasted, and may stay even after some editing.

Watermarking will be applied at the model level, meaning it will be present no matter which Claude product the text comes from.

AI-generated files like images will now contain metadata that shows the file was created and processed by Claude.

Anthropic is working on detection mechanisms to help users find Claude's watermarks and metadata, and will share details soon.

The company is also working to add the watermarking feature to older Claude models.

Having a watermark or metadata doesn't mean the content was completely generated by AI, just that it was processed by Claude models.

For example, human-written text that's entered into Claude for proofreading will carry a watermark after it's processed. The watermark will stay even if a human modifies the text.

Not having a detected mark doesn't mean the content wasn't AI-generated. Images made by Claude could be screenshot and lose their metadata.

Companies have been adopting AI transparency rules due to user concerns.

Over 200 companies, including OpenAI, Meta, Google, and Microsoft, have signed the EU's code of conduct for AI transparency.

The EU transparency rules include labels for AI-generated content, and companies can develop their own labels. AI models must include watermarking or identifying code.

Spotify announced it will clearly label AI-generated artist personas and not include them in personalized music recommendations.

Spotify suggested it's up to artists to disclose if parts of a song were made with AI, though it verifies human artists.

Spotify hopes for transparency in how AI is used in song creation, just like with AI credits.

OpenAI expanded its provenance metadata to cover AI-generated audio, in addition to images.

OpenAI has a tool that lets users check if content was created by its tools.

Canada's federal AI strategy prioritizes transparency to build trust and protect Canadians.

The government is consulting the public on developing its own AI transparency rules, with experts warning that regulations are needed quickly.

Stay informed with accurate news.

Add Global News as a preferred source on Google to see more of our stories.

Get the day's top news stories from Canada and around the world in your email.

Stay updated with the day's top news stories from Canada and around the world in your email.

COPYRIGHT DISCLAIMERS

4English does not own the rights to this content. All rights belong to the owner. 4English only uses this content to help users learn English via the app's features. Contact to request the removal of your content: contact@4englishapp.com.

Pratique d'écoute

Pratique d'écoute