FPPC Technology Panel Recap: Experts Urge Caution on AI Detection, Suggest Content Credentials

The California Fair Political Practices Commission's public outreach and technology program reported June 19 on a panel the agency convened about detecting artificial intelligence in political advertising and possible technical approaches to meet the disclosure requirements of AB 2355.

Eva Hartman, the commission’s executive fellow who organized the series, summarized the April 22 panel that included University of California, Berkeley computer science professor James F. O’Brien, Hive AI CEO Kevin Guo, and Adobe representatives Katie Brook and Andrew Kabak. Hartman said the panel’s principal takeaways were that AI-generated and AI-altered images, audio and video will become harder to detect, detection technology may not keep pace, and no tool guarantees perfect accuracy. "Our team's takeaways from these presentations are as follows," Hartman said, summarizing the need for human investigation and legislative clarity.

Panelists discussed a mix of approaches: academic detection research, proprietary detection products, and Adobe’s content credentials, which embed provenance metadata into an image file. Commissioners and staff said mandatory labeling in statute could reduce the detection burden but leaves enforcement questions. The commission plans another panel in August to study how ethics agencies can adopt AI tools safely and effectively.

Staff noted the issue is shared across ethics agencies nationwide and that the FPPC will study legislative and technical options to meet AB 2355’s disclosure and enforcement obligations. The April panel video is available on the FPPC YouTube page; the commission said it will continue stakeholder outreach and examine the legislature’s appetite for technology-based labeling tools.