OpenAI has released a dataset mapping how people use ChatGPT for work, covering 165 work activities across nearly 100 countries from the first quarter of 2025 through the third quarter of 2026. Published as part of its OpenAI Signals research program, the data is aimed at economists and labour researchers studying how generative AI is used in the economy. The activities range from preparing instructional materials and developing marketing content to programming and analysing business data. They are grouped into Intermediate Work Activities, a category system built on the O*NET occupational framework, which allows comparison with other employment data. Early findings point to less concentration in work-related use over time. Across 74 countries observed in every quarter, the ten most common activities made up an average of 51.1% of classified work messages in the third quarter of 2026, down from 56.0% in the first quarter of 2025. OpenAI says countries appear to specialise in different types of work, though it notes the differences could reflect the kinds of work done locally or uneven access to AI tools. The company also compared ChatGPT usage with the task mix of U.S. jobs and found the two differ substantially. Part of the gap is that about 43% of the employment-weighted task mix for mapped U.S. occupations falls into categories outside the individually reported set, such as moving materials, operating vehicles and maintaining medical records. The dataset covers messages on individual ChatGPT plans and excludes enterprise messages and Codex use, which limits how far it can be read as a picture of all AI use at work. OpenAI says researchers do not read individual conversations. Messages are classified by automated systems after identifying information is removed, results for small groups are withheld, and statistical noise is added. For Australian businesses and professionals, the release changes nothing immediately. It is a research dataset rather than a product change, and it carries no new cost or obligation. The article does not break out figures for Australia, and the aggregated, de-identified data raises no direct Privacy Act issue for businesses. Teams wanting a rough sense of which tasks are common in consumer ChatGPT use could browse the activity categories, but the data does not describe enterprise deployments.