FEC filings reveal 39 US congressional candidates paid for OpenAI subscriptions, with some using AI for political ads despite strict company prohibitions.

Generative artificial intelligence has transitioned from a futuristic curiosity to a standard line item in the budgets of American political campaigns. Recent analysis of Federal Election Commission (FEC) filings reveals that the machinery of the 2026 midterms is increasingly powered by large language models, even when those tools are explicitly forbidden for certain political uses.

The paper trail of AI adoption

A detailed review of federal campaign-finance filings, first highlighted by The Washington Post, identifies 39 congressional candidates who reported payments to OpenAI during the current election cycle. While the number may seem modest, it represents only the visible tip of the iceberg. Political consultants suggest that the actual penetration of AI in campaigns is significantly deeper than what official records indicate.

The financial footprint varies widely across the political spectrum. The Republican National Committee emerged as the largest political spender among the identified groups, reporting approximately ,700 in payments to OpenAI. On an individual level, Representative Mike Lawler of New York has logged roughly ,260 in charges since 2025. Beyond candidates, around 30 political action committees (PACs) and parties also disclosed payments to the company.

Circumventing the political ad ban

The most contentious aspect of these disclosures is the apparent disregard for OpenAI's terms of service. The company has maintained a strict policy prohibiting the use of its tools to generate content for political advertising. However, the FEC filings show that at least two candidates openly disclosed using their subscriptions for advertising purposes.

A third candidate admitted to using unspecified AI tools for more complex tasks, including the creation of synthetic media and the personalization of voter messages. This suggests a growing trend where campaign managers view corporate usage policies as mere suggestions rather than binding constraints. The challenge for OpenAI lies in enforcement; the company does not employ compliance officers to monitor every individual conversation within ChatGPT to ensure no political ads are being drafted.

The invisibility of the subscription

Industry insiders argue that the 39 reported candidates are a fraction of the total users. The primary reason is the low cost and high flexibility of the ChatGPT Plus plan. At per month, a subscription is easily absorbed into vague accounting categories such as 'office supplies' or 'digital consulting,' making it nearly invisible to regulators.

For the most part, people are using it to write their emails, write their ad copy, write their scripts,

This quote from Republican digital strategist Eric Wilson underscores the utility of the tool. When AI is used for internal drafting or scriptwriting, it often bypasses the scrutiny applied to final, public-facing advertisements. Furthermore, many staffers likely use free tiers or personal accounts, ensuring that the financial trail never reaches the FEC filings.

Inconsistent AI guardrails

While OpenAI attempts to maintain a boundary between its technology and political influence, the effectiveness of these guardrails is questionable. Testing conducted by The Washington Post indicated that the AI's refusal to generate political content is inconsistent. In some instances, ChatGPT declined to draft targeted messages for specific demographics or voter emails, only to comply with identical prompts later the same day.

This volatility in AI behavior creates a landscape where some campaigns may struggle to use the tool while others find loopholes through prompt engineering. The result is an uneven playing field where the ability to 'trick' the AI into producing prohibited content becomes a competitive advantage in digital campaigning.

A regulatory vacuum in Washington

The rapid adoption of AI in the 2026 midterms has exposed a significant lag in federal oversight. As of mid-2026, the FEC has yet to finalize binding rules regarding the disclosure of AI-generated electoral content. While the commission has debated the necessity of such rules, the lack of a clear framework allows campaigns to operate in a grey area.

This federal inertia contrasts with the movement in some US states, which have moved faster to implement laws requiring clear disclaimers on AI-generated political content. However, without a national standard, the ability of voters to distinguish between human-authored and AI-generated communication remains limited, fueling skepticism and distrust in political advertising.

Global implications for AI business and governance

For international entrepreneurs and tech firms operating in the US and UK markets, this situation serves as a critical case study in the limits of 'Terms of Service' as a regulatory tool. When a technology becomes essential for operational efficiency—as AI has for political campaigning—corporate bans are rarely effective without technical blocks or government enforcement.

In the United States, the lack of FEC mandates means that AI providers are currently the only entities attempting to police the use of their tools in elections. In the UK, where electoral laws are traditionally more stringent regarding spending and transparency, the challenge remains how to categorize AI subscriptions. If a subscription can produce the work of a high-paid consultant, the traditional definitions of 'campaign spending' may need a total overhaul.

For businesses scaling AI products globally, the lesson is clear: relying on user agreements to prevent misuse in high-stakes environments like national elections is insufficient. As seen in the AI Weekly reports, the demand for efficiency often outweighs the fear of policy violations. Companies must anticipate that their tools will be used in ways they explicitly forbid, necessitating more robust technical safeguards or a shift toward transparent, regulated cooperation with electoral bodies.

FAQ

---

Sources: Diariobitcoin, Aiweekly, Cryptobriefing ·