Data licensing examples

AP and OpenAI: licensing a news archive and election data

A text-archive collaboration followed by a separately announced election-results agreement.

This example helps publishers distinguish historical content from live data feeds and describe the use a license is meant to support.

Deal at a glance

Seller / buyer
Associated Press / OpenAI
Original asset
Selected text archive
First announcement
July 13, 2023
Later asset
Election vote-count results
Later announcement
May 27, 2026
Economics
Not publicly disclosed in reviewed announcements

The original archive agreement

AP announced on July 13, 2023 that OpenAI would license part of its text archive while AP gained access to OpenAI technology and product expertise. The announcement describes exploration of generative AI uses and improving OpenAI systems.

AP: archive and technology collaboration (July 13, 2023)

What the source does and does not establish

The initial source confirms a selected text-archive license, not a license for every AP format. It does not disclose the price, archive date range, file count, exclusivity, delivery requirements or exact division of training and other uses. It supplies no deal-specific revenue or profit figure.

AP: archive and technology collaboration (July 13, 2023)

Later agreement: election results

On May 27, 2026, AP announced that it would provide OpenAI with vote-count results for national, state and local races in major American cities through the 2028 general election. AP described making election information available to users of ChatGPT and OpenAI services. This separate data agreement does not reveal the economics of the 2023 archive license.

AP: OpenAI election-results agreement (May 27, 2026)

Seller lesson: define archive and feed separately

Editorial inference: a historical archive and an updating feed need different descriptions. For an archive, record date coverage, versions and correction history. For a feed, describe update frequency, identifiers and how corrections propagate. Ask a prospective buyer which product they need before building a delivery system.

Seller lesson: specify the intended use

Editorial inference: distinguish permission to train a model from permission to retrieve, display or redistribute content. Ask about attribution, freshness requirements and ongoing support. A general AI collaboration announcement may leave these boundaries unresolved; the proposed agreement should describe them.

What a smaller publisher can take away

Use this case to prepare questions, rather than an earnings benchmark. Describe your original coverage and how you track authorship, corrections and third-party material. Do not assume that a similar topic creates an open route to OpenAI or that access to technology equals cash proceeds.

Sources and verification

Primary sources reviewed October 6, 2026. Historical agreements and later developments are distinguished below.

AP: archive and technology collaboration (July 13, 2023)

AP: OpenAI election-results agreement (May 27, 2026)