Open AI supports California bill requiring ‘watermarks for AI-generated content

Open AI is supporting AB 3211, a California bill that requires labelling of AI-generated content.

The bill requires watermarks in the metadata of AI-generated photos, videos and audio clips, seeking to fight disinformation and promote transparency amid worries about AI’s influence over US general elections.

As most people are still unable to read metadata, the bill also calls for online platforms to label information so that their audiences can clearly understand when content is AI-generated.

AB 3211 follows another California state bill focused on AI that has recently found opposition among Silicon Valley technologists including Microsoft and Open AI, called SB 1047.

This bill requires AI developers to conduct security tests on many of the most advanced AI models that cost more than $100 to develop or a defined amount of computing power, and to hire third-party auditors to assess their safety practices.

It also requires AI software developers working in the state to outline solutions which can turn off AI models in the event of an error.

AB3211 found opposition in April, with a trade group representing Adobe and Microsoft amongst others giant tech firms. However, these doubts were revoked after ammendments were made to the bill.

OpenAI, Adobe and Microsoft are currently part of the Coalition for Content Provenance and Authenticity, which helped create C2PA metadata, an open technical standard providing publishers, creators, and consumers with the ability to trace the origin of different types of media.

California state lawmakers have been trying to introduce several measures designed to ensure that all algorithmic decisions are impartially tested and protect the intellectual property of deceased individuals from exploitation by AI companies.

According to the state legislative database, about 65 legislative proposals were produced during the last legislative season., However, most of the proposals have not been approved.