OpenAI and Google: AI Traceability with C2PA and SynthID
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
OpenAI and Google Unite for Better AI Image Traceability
In response to the proliferation of images generated by artificial intelligence (AI) and their increasing sophistication, OpenAI has announced measures to improve the verification of such content. On Tuesday, the company unveiled two major initiatives to tackle this challenge.
OpenAI has decided to adopt the open standard C2PA (Coalition for Content Provenance and Authenticity), which allows for a signal to be added to the metadata of an image, clearly indicating that it was generated by AI. Founded in 2021, the C2PA is a non-profit organization dedicated to mitigating the harmful effects of AI-generated images on public discourse. This standard is already used by several Google products, but its adoption remains uneven across the industry.
Collaboration with Google for an Invisible Watermark
In parallel, OpenAI is partnering with Google to integrate an invisible watermark called SynthID. This watermark is designed to be difficult to detect and erase, even if malicious actors attempt to remove it through digital manipulations such as screenshots or resizing.
These new protections apply only to images created by OpenAI's products. While they do not affect images generated by other AI tools, they ensure that OpenAI does not contribute to misinformation.
A Verification Tool for the Public
OpenAI has also introduced a public verification tool. This tool will allow users to test whether an image was generated by AI by checking for C2PA and SynthID signals. Initially, this tool will be limited to images produced by OpenAI, but the company plans to expand it to other tools in the future.
A Complementary Approach for Better Resilience
The combination of C2PA metadata and the SynthID watermark aims to enhance the resilience of image provenance. OpenAI emphasizes that "watermarking can be more durable through transformations like screenshots, while metadata can provide more information than a watermark alone." Together, these systems offer increased protection against malicious manipulations.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.