OpenAI’s Usage Policies prohibit use for political campaigning, lobbying, foreign or domestic election interference, or demobilization activities. This article explains what “political campaigning” means under our Usage Policies and provides non-exhaustive examples of campaign-related uses that are either permitted or prohibited. You can read more about our overall approach to elections.
Political campaigning under our policies covers the generation or distribution of scaled campaign messaging (e.g., written, audio, or visual assets) that advocates for or against a candidate, political party, or ballot measure. Advocacy may be direct or implied, and violations of our policies can result in account restrictions or other penalties.
Our steps to limit certain campaign uses are intended to give the public, civil society, and governments more time to adapt to AI. But we support responsible use of AI for civic and campaign work that does not violate our policies.
Campaigns and election-related organizations spend significant time on administrative work, research, planning, data analysis, compliance, accessibility, and operations. When used responsibly, AI can help with those tasks. It can help teams organize information, prepare internal memos, improve accessibility, conduct research, support cybersecurity, and streamline internal workflows. That can free people to spend more time in communities, hearing from voters and generating ideas that meet community needs.
What we allow
Political campaigns, election-related entities, and advocates may use our services for responsible, human-directed work. Examples:
Preparing internal staff memos, briefings, planning documents, and event logistics
Summarizing research, public records, polling, survey results, and other information for internal use
Supporting accessibility, formatting, coding, cybersecurity, compliance, and administrative tasks
Helping a person draft or edit remarks that will be delivered by a person, such as speeches, scripts, or talking points
Brainstorming slogans or narrative themes that someone will deliver
Organizing or analyzing data, such as data used for campaign operations, budgeting, policy development, and trend analysis
What we do not allow
Use of our services must comply with all applicable terms and policies, including Usage Policies regarding privacy, impersonation, election interference, demobilization, and lobbying.
This means that no one may use our services to create campaign messaging content at scale, automate campaign outreach, determine which individuals should receive particular messages, operate public-facing chatbots, or connect messaging content to third-party distribution tools. As is the case for every Usage Policy, our campaign messaging restrictions apply to all users, not only official campaigns or registered advocacy organizations.
Practically, campaigns should ask themselves whether they are generating a workflow that creates or adapts scaled messaging content or automates distribution. If so, the use would violate our policies.
Examples of prohibited uses include creation or automated distribution of:
Campaign emails, texts, direct messages, robocalls, or push notifications.
Posts, comments, messaging responses, or personalizing messaging assets (e.g. ads).
Individualized or segmented campaign messages for distribution, even if a human reviews them before they are sent.
Assessing which individual voter, donor, supporter, or audience should receive which campaign message.
Building public-facing campaign chatbots, agents, or automated systems that interact with voters, donors, supporters, media, or the public.
Creating fake endorsements, fake testimonials, fake voter letters, fake comments, deceptive synthetic media, or other astroturfed activity.
Note: Usage Policies also restrict the use of our tools to evaluate or classify individuals based on their social behavior, personal traits, or biometric data (including social scoring, profiling, or inferring sensitive attributes.