Rep. Kevin Mullin gave Meta until Sept. 18 to explain how it plans to stop deepfakes and AI-generated misinformation from reaching voters on Facebook, Instagram, Threads and WhatsApp ahead of the November midterm elections.

The San Mateo Democrat led 18 Bay Area lawmakers, including Rep. Nancy Pelosi of San Francisco and Rep. Lateefah Simon of Oakland, in a letter sent Aug. 19 to Meta CEO Mark Zuckerberg. The letter calls the company's content moderation "inadequate" and raises concerns about three policy areas: Meta's rollback of third-party fact-checking, its automation of risk review and the unchecked spread of AI-generated content across its platforms.

"American voters have a right to know whether what they are seeing online about candidates and elections is real or fake," Mullin wrote in the letter.

The letter poses nine questions to Zuckerberg. They cover staffing plans and budgets for Meta's election integrity teams and the company's ad disclosure requirements, though the full list of questions has not been published.

Mullin cited specific examples of AI-fueled manipulation already underway. A 22-year-old in India created a fake conservative influencer named "Emily Hart" who pushed anti-abortion and anti-immigration content, amassing millions of views on Instagram before Meta removed the account. Deepfake videos of candidates speaking to the camera have also circulated in Senate races in Texas and Georgia, according to the letter.

Meta has not commented, KQED reported.

State Sen. Aisha Wahab, D-Hayward, said she would file a Federal Election Commission complaint over a "surge of suspicious accounts, bot-like activity" on Meta's platforms during the Aug. 18 special election to fill the congressional seat formerly held by Eric Swalwell, as KQED reported. Wahab's complaint and Mullin's letter are separate actions, but both target Meta's handling of election-related content.

Separately, a California AI transparency law co-authored by state Sen. Josh Becker, D-San Mateo, took effect Aug. 1. That law requires AI-generated content to carry embedded metadata about its origin; large social media companies, including Meta, must detect and disclose that data starting in January 2027, with violators facing a $5,000 civil penalty per violation.

Meta has until Friday, Sept. 18 to respond to Mullin's nine questions.