The San Francisco County government meeting on July 4, 2025, focused on several significant legislative proposals aimed at addressing the challenges posed by artificial intelligence (AI) and deep fakes. Key discussions revolved around four bills currently progressing through the state assembly and senate, which seek to enhance the integrity of information and protect the electoral process.

The first bill, Assembly Bill 3211, introduced by Assemblymember Buffy Wicks, mandates that AI-generated content must include an invisible watermark to indicate its origin. This proposal also requires manufacturers of phones and cameras to embed similar watermarks in authentic media, aiming to clarify what is real and improve the overall information environment.

The second bill, Assembly Bill 2655, sponsored by Assemblymember Mark Berman, targets the circulation of deep fakes, particularly during election periods. It seeks to impose restrictions on deep fakes related to elected officials and election infrastructure, with a focus on minimizing any potential infringement on First Amendment rights.

A related measure, the "Deep Fake Free Campaigning" bill, also from Assemblymember Pellerin, addresses the use of deep fakes in offline media, such as billboards and print advertisements, ensuring that similar restrictions apply beyond digital platforms.

Lastly, Senate Bill 1228, introduced by State Senator Padilla, aims to enhance online security by requiring platforms to verify the identities of influential users, ensuring they are real individuals rather than bots or foreign interference actors.

During the meeting, there was a discussion about the possibility of the San Francisco Board of Supervisors passing a local version of the 2019 deep fake law, AB 730. However, it was recommended that the board wait until the outcomes of the current bills are clearer, particularly AB 2655.

The meeting also touched on the importance of preparing election officials for AI-related threats, highlighting a recent tabletop exercise in Arizona that involved over 200 local election officials. A report from this exercise provides a checklist for identifying and responding to potential AI threats in elections.

Overall, the discussions underscored the county's proactive approach to navigating the complexities of AI and misinformation, with a focus on safeguarding democratic processes and enhancing public trust in information. The next steps will depend on the progress of the proposed bills in the state legislature.