OpenAI Expands AI Content Provenance Since 2024, Backing EU Transparency Code

2 articles · Updated · OpenAI · Jun 11

OpenAI said it will strengthen content transparency across its products, improve verification tools and support Europe’s push for a more trustworthy AI ecosystem.
Since 2024, the company has added C2PA metadata to DALL-E 3 images, layered SynthID watermarks onto images from ChatGPT, Codex and its API, and launched a public verification site for supported images.
OpenAI said provenance matters because it can show where content came from, how it was edited and whether it is authentic, helping counter disinformation and protect election integrity.
The company also stressed current limits: metadata can be stripped, watermarks can degrade and labels only work where users see them, making broader industry cooperation essential.
That stance extends OpenAI’s European AI governance push after it became the first U.S. company in 2025 to sign the EU’s General-Purpose AI Code of Practice.

Sources

OpenAI11h ago

OpenAI Strengthens AI Content Transparency, Supports European Trustworthy AI Ecosystem

t-systems.com11h ago

I am AI. And I believe Europe is underestimating its greatest opportunity – T-Systems

With C2PA's security flaws now public, are AI watermarks a false promise for stopping deepfakes and misinformation?

If social media strips AI watermarks, how can we ever truly verify the origin of viral online content?

AI Content Provenance in 2024: How OpenAI, Google, and Global Regulations Are Shaping the Future of Digital Authenticity

Overview

OpenAI is responding to new regulations like the AI Act by strengthening its approach to AI content transparency. The company now uses both watermarking and metadata systems—specifically, C2PA for detailed image metadata and SynthID for durable watermarking. These two systems are designed to cover each other's weaknesses, making it harder for AI-generated content to go undetected. By combining these methods, OpenAI creates a more robust way to trace the origin of digital media. However, their current verification tool only works for images made by OpenAI products, showing that universal solutions are still in development.

...