Deepfakes, or AI-generated synthetic media, are causing increasing concern as the 2020 elections approach. Deepfakes can make people appear to say or do things they did not actually say or do. As deepfake technology becomes more advanced and easier to use, some experts fear it could be deployed to spread misinformation or generate “viral” content during the 2020 campaign season.

Lawmakers and researchers say deepfakes pose a threat to the democratic process. Deepfakes could be used to generate false audio or video of candidates saying inflammatory or offensive things in the final days of a close election. They could also be used to spread misinformation about voting procedures or results on social media.

There are concerns that the public and media may have trouble distinguishing deepfakes from real content. As deepfake technology improves, the fakes are becoming very realistic. And people tend to believe and share shocking or sensational content, even if it’s not real.

Social media platforms say they are working to curb the spread of misinformation, including deepfakes. But detecting them is challenging, and critics argue the companies have not done enough. Legislation has also been proposed to criminalize some uses of deepfake technology, though it’s unclear if any bills will be passed into law before 2020.

Some experts argue that the threat of deepfakes is overblown and that countermeasures like media literacy campaigns and fact-checking can help limit their impact. But others say that as deepfakes become more advanced and automated, the scale at which they can spread on social media will overwhelm current countermeasures. They say policymakers need to take the threat far more seriously to protect the integrity of the 2020 elections.

The debate around deepfakes is a microcosm of the broader debate around AI and elections. Optimists believe that AI can improve civic engagement and even help combat misinformation. Pessimists worry that malicious actors will co-opt AI technologies to undermine democracy. The truth likely lies somewhere in the middle. AI is a tool, and whether its impact on elections is positive or negative depends entirely on how it’s used.