AI & Democracy
AI can help strengthen democracy, or help collapse it
AI and social media can help voters find accurate information, help public officials serve the public, and help more people take part. But that depends on trust, accurate information, and neutral pipes that don't manipulate the algorithms for partisan purposes.
Trustworthy AI & Elections · The statement is open for signatures, for 2026 and beyond
Sharing this? Sample posts and emails are in the media kit.
Why this matters
AI can strengthen democracy, or wear it down
How AI can strengthen democracy (examples)
- Help voters find accurate information about candidates, ballot measures, and how to vote.
- Help election officials answer voters’ questions and serve the public.
- Lower the cost of taking part, in more languages, for more people.
- Make it easier for people who disagree to find common ground, at a scale town halls can’t reach.
How AI can collapse democracy (examples)
- Synthetic video and audio of real candidates that voters can’t tell from the real thing.
- AI agents that contact voters without saying they’re AI, who sent them, or share misleading information.
- Ranking and recommendation systems that shape what voters see, can be manipulated for partisan or anti-democratic purposes, with no way for the public to check them.
- Chatbots that give wrong answers about when, where, and how to vote.
Voters have long relied on clear rules in broadcast political advertising: sponsors identify themselves, candidates stand by their ads, and the public can see who paid. Social media brought only partial, mostly voluntary versions of those rules, and led to various declines, social fragmentation, heightened polarization beyond what actually exists, and the spread of disinformation. AI systems and recommendation algorithms now shape much of what voters see, and more and more they talk to voters directly. The same basic expectations of honesty, transparency, and fairness should apply to them.
Our campaign
Trustworthy AI & Elections
A cross-partisan sign-on statement: five principles for AI and platforms in elections, and a call for companies to show the public, on the record, how they meet them. For 2026 and beyond. It endorses no candidate, party or bill.
- Principles anyone can support
- Companies show the public
- Officials of every party ask, on the record
- Voters can trust what they see
The five principles
- 1. Authentic and clearly labeledReal content from campaigns. AI clearly labeled. No deceptive deepfakes.
- 2. Transparent fundingPolitical ads show who paid, including through intermediaries and influencers.
- 3. Fair and transparent platformsSame rules for every candidate and viewpoint, open to independent checks.
- 4. Accurate voting informationAI points voters to official sources on when, where and how to vote.
- 5. Accountability in publicCompanies report threats and fixes before and after each election.
The wider ecosystem
AI & Democracy Directory
438 organizations, campaigns and ballot measures across the US working on AI and democracy, from AI safety and kids’ online safety to election integrity and data centers. It maps the field; being listed doesn’t mean a group is part of our campaign.
How to help
What you can do
Voters and citizens
- Get campaign updatesNewsletter
- Get news from several reliable sourcesAllSidesNews Literacy Project
- Get voting information from official sourcesCan I VoteYour election office
Public officials
Officials of every party can ask AI and platform companies for written, public answers, use the oversight powers they already have, and set up fast channels for correcting false voting information.
What the statement asks of officials →Organizations
Bring your organization in as an endorser, or help shape what comes next.
Email ai@democracyweek.net →Resources
Key resources
Key sources on each topic. Open a topic to see them.
Political ad librariesWhere to see who’s running political ads7
- Meta Ad Library and Ad Library Report · MetaSearchable archive of political and issue ads on Facebook and Instagram, with spending and funders.
- Google political ads policy and Ads Transparency Center · GoogleElection ad verification, “Paid for by” disclosures, and a public ad archive.
- Political Ads Disclosure · XX’s disclosure reports for US political ads.
- Snap Political Ads Library · Snap Inc.Public library of political and advocacy ads on Snapchat.
- TikTok: policies on paid ads · TikTok NewsroomTikTok bans paid political ads, including paid creator content.
- LinkedIn's Ad Library · LinkedIn (Microsoft)General ad library. LinkedIn doesn’t accept political ads.
- FCC Public Inspection Files · Federal Communications CommissionBroadcast and cable stations’ political files: the long-standing model for ad disclosure.
Who pays for political messagesCampaign finance data2
- FEC campaign finance data · Federal Election CommissionOfficial fundraising and spending data for federal candidates and committees.
- OpenSecrets · OpenSecretsNonpartisan data on money in politics, including outside spending.
Deepfakes and AI disclosure lawsState and federal rules, and court limits6
- Tracker: State Legislation on Deepfakes in Elections · Public Citizen, Updated Oct 5, 2026State-by-state bill status; table shows about 33 enacted laws.
- Artificial Intelligence (AI) in Elections and Campaigns · National Conference of State LegislaturesState statutes on deepfakes in political messaging; prohibitions vs. disclosures.
- FCC Makes AI-Generated Voices in Robocalls Illegal · Federal Communications Commission, Feb 8, 2024Declaratory Ruling FCC 24-17: AI voices are 'artificial' under the TCPA.
- Protect Elections from Deceptive AI Act (S.1213) · Congress.gov, Introduced Mar 31, 2025Bipartisan Senate bill on deceptive AI media of federal candidates.
- FAIR Elections Act of 2026: summary · Office of Sen. Jeff MerkleySenate bill on AI-generated vote-suppression content and voter rolls.
- Babylon Bee 1, California 0: Court Strikes Down Law Regulating Election-Related AI Content · American Enterprise InstituteOn the court ruling that struck down California’s election deepfake law.
AI chatbots and accurate voting informationHow chatbots answer voting questions5
- #TrustedInfo2026 · National Association of Secretaries of StateSecretaries of state campaign pointing voters to election officials.
- Can I Vote · National Association of Secretaries of StateNonpartisan hub for registration and polling place information, by state.
- Seeking Reliable Election Information? Don't Trust AI · AI Democracy Projects (Proof News and Institute for Advanced Study), Feb 27, 2024Experts rated about half of five chatbots’ election answers inaccurate.
- Generating Confusion: AI chatbot answers on voting with a disability · Center for Democracy & Technology, 2024Found errors in chatbot answers on accessible voting rules.
- What Voters Think Ahead of the 2026 Midterms: Election Information, Confidence, and Concerns · Bipartisan Policy Center, 2026Most voters, across parties, worry about AI-driven election disinformation.
Platform transparency and researcher accessOpening systems to independent checks5
- Better Access: Data for the Common Good · Knight-Georgetown Institute, Nov 2025Framework for independent researcher access to public platform data.
- Better Feeds: Algorithms That Put People First · Knight-Georgetown Institute, Mar 4, 2025Expert guidance on recommender-system design and transparency.
- Coalition for Independent Technology Research · Coalition for Independent Technology Research, Oct 12, 2022Membership group defending independent research on technology and society.
- Integrity Institute · Integrity InstituteThink tank of trust-and-safety professionals from 80+ tech companies.
- The Stanford Internet Observatory is being dismantled · Platformer, Jun 2024SIO scaled back election research after lawsuits and political pressure.
Content provenanceStandards for showing where media came from2
- C2PA (Coalition for Content Provenance and Authenticity) · C2PAOpen standard (Content Credentials) for showing where media came from and how it was edited.
- Content Authenticity Initiative · Content Authenticity InitiativeCross-industry effort and open-source tools for Content Credentials.
What the companies say they’re doing in 2026Company election pages6
- Anthropic: Preparing for the 2026 US midterms · Anthropic, Oct 14, 2026Democracy Works voting info, election-night results, red-teaming, enforcement.
- Anthropic: 2026 usage policy update · Anthropic, Oct 8, 2026Elections section renamed; blanket ban on personalized campaign targeting removed.
- OpenAI: Election information and safeguards in 2026 · OpenAI, May 2026Election results sourcing, watermarking, and no political ads this cycle.
- Google: Supporting the 2026 US midterm elections · Google, Sep 2026Official voting info in Search and Gemini, SynthID, security tools.
- Meta: Preparing for the 2026 US midterm elections · Meta, Feb 2026Final-week new-ad block, AI disclosure labels, Community Notes.
- Microsoft: Helping voters navigate election information in the age of AI · Microsoft On the Issues, Sep 15, 2026'Check. Recheck. Vote.' AI literacy campaign with NASS and NASED.
Past commitments, and how they held upThe 2024 pledges, reviewed4
- A Tech Accord to Combat Deceptive Use of AI in 2024 Elections · Munich Security Conference, Feb 16, 2024Voluntary eight-commitment framework signed by major AI and platform firms.
- Tech Companies Pledged to Protect Elections from AI. Here’s How They Did. · Brennan Center for Justice, Feb 13, 2025Assessment of Munich accord follow-through (Ahmed, Doyle, Harris, Norden).
- Bipartisan election officials, security experts and technologists call for AI election safeguards · Issue One, May 2024Letter to Senate leaders backing three bipartisan AI election bills.
- Launch of the AI + Election Security Coalition · The Future US (via ACCESS Newswire), Feb 15, 2024The Future US and Issue One coalition and open letter.
Recent oversightHearings and investigations this fall4
- FTC issues subpoenas to Anthropic and OpenAI over safety incidents · CBS News, Oct 15, 2026FTC civil investigative demands on safety incidents, privacy, safeguards.
- Anthropic, OpenAI, Google and Meta executives testify at NYC Council AI hearing · CNBC, Oct 5, 2026
- SpaceX agrees to testify to NYC Council about AI safety after receiving subpoena · CNBC, Oct 13, 2026
- SpaceX AI policy leader grilled by NYC Council on Grok safety, sexualized images · CNBC, Oct 16, 2026
Perspectives from across the spectrumCenter-right and bipartisan views3
- AI and Elections: What to Watch for in 2026 · R Street Institute (Chris McIsaac), Jan 27, 2026Center-right view favoring disclosure over bans.
- AI Chatbots Are Reshaping Political Persuasion · American Enterprise Institute (John Bailey, Julia Torres), Dec 22, 2025Research showing chatbots can shift political attitudes.
- Preparing for AI & Other Challenges to Election Administration · Bipartisan Policy Center, May 2026Practical guidance for election officials.
Where voters get informationSurvey data on chatbots and social media2
- Americans and AI 2026: Chatbots, Smart Devices and Views on Impact · Pew Research Center, Jun 17, 2026Which AI chatbots Americans use, and for what.
- Social Media and News Fact Sheet · Pew Research Center, 2025 surveyWhich social platforms Americans regularly get news on.
Stay in the loop
Follow the campaign
News on the statement, and how to take part.
Spread the word
Help promote this event
Sample posts, emails, logos, and graphics for the Trustworthy AI & Elections statement, ready to copy and make your own. A personal note to a few people who should be there does more than any post.