USA: California’s anti-deepfake bills are a good start, but more is needed
The California Senate and Assembly recently passed three bills limiting the use of deepfakes during election periods. These bills, which await the governor’s signature, are a positive first step – but Reporters Without Borders (RSF) calls for this regulation to be strengthened and adopted at the federal level.
California, home to the world’s largest artificial intelligence (AI) companies, is attempting to regulate deepfakes for the first time. Like similar laws recently passed in other states, the three new bills aim to limit the creation and dissemination of certain types of deepfakes. Passed at the end of August, the legislation would immediately take effect once signed by the governor.
Yet unlike other deepfake laws, California’s bills are specifically restricted to election periods. The first bill, AB 2839, prohibits the intentional dissemination of digitally manipulated audiovisual content – including deepfakes – that misrepresent a candidate, elected official or an elections official in the periods before and after elections. The second text, AB 2655, requires digital platforms to identify and remove this type of content during election periods, and label any deepfakes likely to compromise the vote. The third bill, AB 2355, requires all election communications that feature content generated or modified by AI to include a clear statement indicating its synthetic nature.
"Deepfakes featuring public figures, such as journalists or election candidates, can garner millions of views on social networks. They disrupt the public’s relationship with facts, undermine trust in information, and harm democracy. They must be addressed on all fronts: political, legal and technological. The bills currently being passed in California are a necessary but inadequate first step in this direction. It is essential to strengthen these regulations and implement them at the federal level as quickly as possible.
Deepfakes: a threat to democracy
Democratic candidate Kamala Harris, victim of a deepfake video that went viral on X (formerly Twitter) in July, was targeted by another disinformation campaign in September involving an AI-generated fake news report. Deepfakes that impersonate journalists and legitimate media outlets not only spread false information, but erode public trust in journalism.
A necessary – but insufficient – first step
Limiting the dissemination of deepfakes during election periods is just one part of RSF's recommendations. These bills must be expanded to be fully effective.
First and foremost, the content banned by these bills should not be limited to deepfakes representing candidates and election officials: at the very least, these bills should ban any deepfake that falsely portrays a public figure – particularly journalists.
RSF also recommends introducing criminal penalties for any person or organization that intentionally publishes or disseminates deepfakes to mislead the public, particularly when they imitate someone without their consent. This measure has already been adopted in France and Italy.
The bills must also hold developers and suppliers of generative AI systems accountable for preventing these harmful deepfakes. Lastly, RSF encourages the adoption of technical standards that verify the authenticity of audiovisual content, which will boost trust at a time when deceptive content is proliferating.
Stronger legislation is needed at the federal level
Two bills, similar to those adopted in California, have been introduced to the US Senate, but have not yet been scheduled for debate. The first, S.2770, introduced in September 2023, prohibits the distribution of deepfakes representing candidates or elected officials. The second, S.3875, introduced in March 2024, requires election advertisements to disclose when artificial intelligence is used. RSF has also expressed support for the COPIED Act, a bill allowing content creators to attach content provenance information to their works.
Stronger, more coordinated national legislation – accompanied by international cooperation with other countries – is vital in combating the threat to reliable information currently posed by deepfakes.
