Variety via Reddit

OpenAI Embeds Live AP Election Returns in ChatGPT

openai generative ai regulation elections data-licensing cybersecurity

Key insights

  • AP will supply OpenAI with live US and Brazilian election results embedded directly in ChatGPT through the 2028 general election.
  • The deal extends a 2024 content-licensing arrangement, making ChatGPT the first major AI assistant to serve live vote counts.
  • OpenAI is separately offering election-security products to voting-system manufacturers and briefing state election officials ahead of 2026 midterms.

Why this matters

OpenAI embedding live AP election data inside ChatGPT establishes a precedent for AI assistants becoming primary election-night information conduits, bypassing traditional media and web search at scale. The pairing of real-time authoritative data with a generative model raises novel accuracy-guarantee questions: AP's data reliability does not automatically prevent ChatGPT from misframing or contextualizing around correct numbers. OpenAI's simultaneous move into election security briefings and product sales to voting-system manufacturers signals a land-grab for government trust relationships that will shape regulatory treatment of AI in civic infrastructure for years.

Summary

OpenAI locked in a multi-year deal with AP to surface real-time election results inside ChatGPT, covering US races and Brazilian elections through 2028. The deal extends a 2024 content licensing arrangement and marks the first time ChatGPT will show live vote counts as returns arrive. AP's century-old infrastructure becomes OpenAI's authoritative feed at the exact moment AI electoral misinformation is under maximum scrutiny. Essentially: (AP, OpenAI) are betting that pairing a trusted wire source with mass-consumer AI beats a fragmented information environment. - AP covers US national, state, local, and Brazilian races embedded natively in ChatGPT through 2028. - OpenAI is separately briefing state election officials and offering Codex Security tools to voting-system manufacturers before 2026 midterms. ChatGPT serving tens of millions on election nights makes result framing a new category of public trust problem.

Potential risks and opportunities

Risks

  • If ChatGPT misframes AP-correct numbers during a close 2026 midterm race, AP bears reputational risk for accuracy failures that were actually model-layer problems.
  • Voting-system manufacturers accepting OpenAI's Codex Security and Trusted Access for Cyber products create a shared attack surface: a compromised OpenAI tool could affect multiple certified election vendors simultaneously.
  • A ChatGPT outage or API failure on election night 2026 could suppress voter confidence in results for millions of users relying on it as their primary returns source.

Opportunities

  • Competing AI platforms (Google Gemini, Perplexity, Meta AI) face pressure to sign similar data partnerships with Reuters, Edison Research, or Decision Desk HQ before the 2026 midterms.
  • Election data providers outside AP (Edison Research, Decision Desk HQ) gain negotiating leverage with AI platforms now that AP has set a public precedent for this category of deal.
  • Civic-tech and election-integrity firms (MITRE election security practice, Galois) may see increased inbound from state officials newly aware of AI vendor security products following OpenAI's NASS briefing.

What we don't know yet

  • Revenue terms of the AP-OpenAI deal are undisclosed; unclear whether AP receives per-query fees, a flat license, or equity-style compensation.
  • Whether ChatGPT vote-count responses will include AP's race calls or only raw numbers, and how the model handles contested or uncalled races at the time of query.
  • No public reporting on whether Brazil's Superior Electoral Court (TSE) approved or was consulted about including Brazilian election data in the arrangement.