Protecting New York voters from AI-generated election misinformation

AI-generated misinformation is spreading

AI tools have become inexpensive, widely available, and easy to use. Although AI tools can be used for a variety of beneficial purposes, they can also be misused. As a result of advancements in AI technology, bad actors are able to create misinformation materials faster than ever before. It has been said that a lie can travel halfway around the world while the truth is still putting on its shoes, and that quote has been made all the more accurate in the age of social media. Once an AI-generated image or video is posted on social media, it can be shared rapidly, and is then able to spread like wildfire. By the time the misinformation is unmasked, too often the damage has already been done.

What is a deepfake?

With AI technology, users can easily create lifelike images, video, or audio with a simple text prompt. 

A deepfake can appear to be a real photograph, video, or audio recording of a real person. It can show someone making statements that they never made or doing something they never did. 

Bad actors can try to misuse deepfakes to influence elections

Deepfakes can be created for a variety of purposes. In the context of elections, deepfakes can be exploited by bad actors to try to influence voter viewpoints or behaviors. For example, these materials may aim to deceive voters about candidates; eligibility to vote; and where, when, or how to vote. These deepfakes may try to get you to vote (or not) for a particular candidate, or prevent you from voting at all.

For example, deepfakes might:

• Impersonate a government official providing false instructions about voting
• Try to intimidate you into not voting by threatening something that would never happen 
• Include fake audio clips in phone calls, such as robocalls, to your phone

It can be very difficult to differentiate deepfakes from authentic media. Shortcomings in the technology might cause deepfakes to have “clues” showing they are fake, but remember that the absence of such clues is not a guarantee that something is real. The technology is constantly improving, and flaws are becoming less common. Clues to spot deep fakes might include:

• unnatural appearance or body movements
• unusual lighting, shadows, or audio
• content that is not reported on by established news outlets

Deepfakes are often shared, and can even go viral, across the web and social media. 

Be careful when using chatbots

Chatbots are AI programs that can respond to user requests and questions. While useful for many purposes, chatbots sometimes provide incorrect answers, including to questions about voting and elections. If you rely on the information provided to you by a chatbot, you could end up at the wrong poll site or on the wrong day. Of course, be on your guard against any kind of misinformation – whether it’s a postcard, an email or text, or lawn signs you drive past.

For information about voting, consult the New York State Board of Elections or the local board of elections for your city or county. To find your local board of elections, you can use the county board roster tool. For basic poll site, registration, and ballot tracking information, voters can visit the State Board of Elections’ voter lookup tool.

Top tips to protect yourself

1. Consult New York state and local boards of elections for information and questions about where, when, or how to vote, as well as your eligibility to vote. The OAG will operate an Election Protection Hotline, which voters can contact to ask questions and file complaints during the voting period.
2. Do not rely on chatbots to answer questions about elections or voting. 
3. If you encounter social media content that you think is a deepfake, report it on the platform. 
4. If you come across information about elections or voting from unofficial sources, verify it with official sources and encourage others to do the same. 
5. Be skeptical of images, audio, and videos circulating online about the elections, candidates, or their campaigns. If you see emotionally charged, sensational, and surprising content, be especially cautious. Remember that it may be fake.
6. Do not share content that you suspect may be fake unless you can verify that it is real.
7. Deepfakes can leave clues showing they are fake, but the technology is getting better all the time and fakes are harder to spot. The absence of clues is not a guarantee that the content is real.
8. Although sometimes there are disclosures that content is AI-generated, often AI-generated content is not labeled as such. 
9. There are tools for detecting AI-generated content, but these are not always accurate. 
10. If you receive a phone call with information about voting or elections, be mindful that the information may be inaccurate or fake. Remember to verify information with official sources.
11. Report misinformation about elections or voting, particularly misinformation concerning the time, place and manner of voting, to our office.