Federal Judge Blocks California’s Laws Regulating AI in Political Campaigns, Citing Free Speech Concerns

A federal judge has struck down two California laws aimed at curbing the use of artificial intelligence (AI) in political campaigning, ruling that the measures infringe on constitutional free speech protections and conflict with federal internet regulations.

Senior U.S. District Judge John Mendez, appointed by President George W. Bush, issued the decision on Friday, halting enforcement of Assembly Bill 2839 and Assembly Bill 2655. The ruling comes as California prepares for what experts are calling the nation’s first AI-dominated election cycle in 2026.

Assembly Bill 2839, signed into law by Governor Gavin Newsom in 2024, prohibited the use of AI-generated content—such as deepfakes and digitally altered media—in political advertisements during the 120 days preceding an election. The bill was challenged in court by a group of content creators, including internet personalities and the satirical news outlet The Babylon Bee, who argued that the law amounted to censorship of protected speech.

In his ruling, Judge Mendez acknowledged the risks AI-generated disinformation poses to election integrity but emphasized that state governments cannot regulate political speech in a way that suppresses constitutionally protected expression.

“Deepfakes and artificially manipulated media undoubtedly raise serious concerns for democracy,” Mendez wrote. “However, the answer to these challenges does not lie in preemptive censorship. Just as the government cannot dictate the bounds of satire or humor, California cannot sanitize political speech under the guise of protecting electoral integrity.”

The court also struck down Assembly Bill 2655, which required digital platforms to monitor and remove AI-generated political content deemed misleading or deceptive. That law was challenged by tech companies, including social media platforms X (formerly Twitter) and Rumble, who argued it violated Section 230 of the federal Communications Decency Act—a law that shields online platforms from liability for third-party content.

Mendez had previously issued a preliminary ruling in favor of the tech firms, stating that AB 2655 placed an unconstitutional burden on digital platforms and directly conflicted with federal law. The final ruling on Friday made that judgment permanent.

The two laws were part of a broader initiative by California lawmakers to prevent what they saw as an impending flood of AI-generated disinformation leading up to future elections.

When signing the bills in September 2024, Governor Newsom warned about the dangers of rapidly advancing generative AI technologies being used to distort facts and erode public trust in democratic institutions.

“It’s critical that we ensure AI is not weaponized to mislead voters and sow confusion in our political process,” Newsom said at the time. “The future of our democracy depends on the public’s ability to trust what they see and hear.”

Supporters of the legislation pointed to the increasing ease with which bad actors—foreign governments, conspiracy theorists, or rogue political campaigns—can now generate convincing false videos, audio clips, and images of public figures. In a fact sheet promoting AB 2839, lawmakers described a near-future scenario where voters might receive robocalls featuring synthetic voices impersonating public officials, or see videos falsely depicting candidates accepting bribes.

“This is California’s first generative AI election,” the fact sheet warned. “Disinformation driven by AI will pollute our information ecosystem like never before.”

The bill sought to limit the spread of manipulated political media during critical election periods—120 days before and 60 days after elections for material targeting voting procedures or election officials. It also mandated disclosure when a candidate’s campaign used AI-generated content to depict themselves doing or saying something they never actually did.

Additionally, the law offered an expedited legal pathway to seek injunctive relief against those distributing such content.

Despite these guardrails, Judge Mendez found the legislation too sweeping, warning that attempts to preemptively restrict political speech—even if well-intentioned—clash with First Amendment protections.

“The government cannot be the final arbiter of what constitutes misinformation in a political context,” he noted. “Such authority is inherently susceptible to abuse and fundamentally at odds with the democratic values it seeks to protect.”

Critics of the laws applauded the ruling, calling it a win for free expression in the digital age.

“This decision affirms that the government cannot silence speech simply because it finds it offensive or potentially misleading,” said an attorney representing The Babylon Bee. “Satire, parody, and political commentary have always had a place in American discourse—even if delivered in new, AI-driven formats.”

With the 2026 elections on the horizon and generative AI tools becoming more powerful and accessible, legal experts predict similar battles will emerge across the country as states attempt to navigate the balance between protecting elections and preserving free speech rights.

For now, California’s attempt to lead the charge on AI regulation in political campaigning has hit a significant roadblock, with broader implications for how democracy functions in the digital age.