The FBI today announced that they are positive Russian and other state actors are using artificial intelligence AI to attempt to manipulate the United States presidential election. I do plan to go into each of these topics individually over the next several days, as it is vitally important to our national security.

Russian and other adversarial entities are increasingly using artificial intelligence (AI) to interfere with U.S. elections through a variety of sophisticated tactics designed to manipulate public opinion, sow discord, and undermine confidence in democratic processes. AI’s ability to rapidly generate and distribute content, coupled with its data analysis and personalization capabilities, has made it a powerful tool in election interference. Here are some of the ways adversaries are using AI to interfere with U.S. elections:

AI-Generated Disinformation Campaigns

One of the most prominent ways AI is being used to interfere with elections is through the creation and dissemination of disinformation—false or misleading information intended to deceive the public. AI tools, such as natural language generation models, are used to produce large volumes of fake news articles, social media posts, and comments that spread misinformation on key political issues, candidates, or election integrity.

Deepfakes: AI-generated deepfake videos and audio can make it appear as though political figures are saying or doing things they never did. These deepfakes are used to manipulate public perception, smear candidates, or create confusion about important issues during an election cycle. For example, an adversary might create a deepfake video of a candidate making inflammatory statements, which can go viral before being debunked, leaving a lasting impression on viewers.

Fake Social Media Accounts: Adversaries use AI to create fake profiles that look and behave like real people on social media platforms. These AI-generated accounts can mimic human interactions, spread disinformation, and amplify divisive content in a coordinated manner, making it harder to distinguish between authentic grassroots movements and orchestrated manipulation.

Amplifying Social Division

AI is also used to amplify existing divisions within U.S. society by targeting specific groups with tailored disinformation. Russian entities, for example, have used AI to exploit algorithmic recommendations on platforms like Facebook, Twitter, and YouTube, where AI systems recommend content based on users’ preferences and behavior.

• Micro-targeting: AI systems analyze vast amounts of data on social media users, including their political affiliations, interests, and online behavior. Adversaries use this data to micro-target individuals or groups with highly personalized political ads, posts, or content designed to manipulate their views. For instance, AI could be used to target specific demographics with misleading information on voting procedures or false claims about voter fraud, thereby eroding trust in the election process.
• Polarizing Narratives: AI-driven bots and trolls are used to push polarizing content that deepens existing divisions on issues like race, immigration, healthcare, or gun control. By stoking both sides of contentious issues, adversaries create the perception of widespread conflict and drive wedges between different social and political groups, making it harder for the electorate to unite around shared democratic values.

3. AI-Driven Bot Networks

AI is heavily used to power bot networks—automated accounts that can produce and share content at high volumes. These bots are used to:

• Flood social media platforms: AI-generated bots can post vast amounts of content on platforms like Twitter and Facebook, making it seem as though certain topics or narratives are more popular than they really are. They also engage with human users by liking, sharing, or commenting on posts to give the appearance of organic engagement.
• Hashtag Hijacking and Trend Manipulation: AI bots can hijack hashtags or manipulate trending topics by coordinating posts to make certain issues appear more relevant or urgent than they are. This tactic can shift media focus, distract from important election issues, or elevate divisive topics to the national conversation.

4. Automated Misinformation on Voting Procedures

Adversaries may use AI to spread disinformation about voting procedures, such as incorrect information on polling locations, dates, or requirements for voter registration. This could take the form of AI-generated content disseminated across social media platforms, emails, or text messages aimed at confusing or disenfranchising specific voter groups. For example, AI-powered bots may send false information that leads people to believe they can vote online (when they cannot), or that specific polling places are closed or relocated.

5. AI-Powered Cyberattacks

In addition to disinformation, adversaries are using AI in cyberattacks designed to disrupt election infrastructure or steal sensitive voter data. AI enables more sophisticated forms of hacking, such as:

• Phishing campaigns: AI can generate more convincing and personalized phishing emails that trick election officials or campaign staff into revealing passwords or downloading malware. AI helps tailor these phishing attempts based on previous data or publicly available information about the targets, making the attacks harder to detect.
• Automation of Network Intrusions: AI is used to automate network intrusions into election systems, such as voter databases or websites that provide election information. AI tools can quickly scan for vulnerabilities, breach systems, and extract data, while potentially masking the attack to avoid detection.

6. AI and Voter Manipulation through Hyper-Targeting

Adversaries can use AI to identify and manipulate key segments of voters through behavioral analysis. AI can analyze users’ preferences, social media activity, and web history to identify undecided voters or those susceptible to persuasion. By leveraging this data, AI can generate personalized disinformation designed to influence specific voter groups or suppress voter turnout for certain populations.

For example, Russian operatives used social media platforms in the 2016 U.S. election to disproportionately target African American voters with disinformation campaigns that aimed to depress turnout. AI’s ability to scale and fine-tune these efforts makes such interference even more efficient.

7. Adversarial AI Attacks on AI Systems

Adversaries may also attempt adversarial attacks on AI systems used by the U.S. government or campaigns. These attacks involve manipulating the input data or exploiting weaknesses in machine learning algorithms to cause misclassification, bias, or failure in AI-driven systems, such as those used for content moderation, cyber defense, or electoral security.

In response, the U.S. government and tech companies are developing stronger safeguards to detect and counter AI-driven disinformation and cyber threats. This includes enhancing AI detection systems to identify fake accounts, deepfakes, and coordinated bot networks, and creating more robust cybersecurity defenses to protect election infrastructure.

However, the rapidly evolving nature of AI and its accessibility means that adversaries will continue to find new ways to leverage the technology to interfere with U.S. elections.