OpenAI backs voluntary bill to track how AI is used at work

Why it matters

OpenAI supports a federal effort to collect voluntary, aggregated data on AI use at work while publishing research based on its own products. The gap between product-use patterns and actual job outcomes is exactly what better workforce data could help measure.

OpenAI policy chief Morgan Dwyer said at a New York City Council hearing on October 5th that the company supports a federal bill to collect data on AI use at work and plans to make its own job-impact analysis public. Dwyer said OpenAI's economic research team was studying the question, but offered no number when asked for one.

The bill, the Workforce Transparency Act of 2026, was introduced in the Senate on April 30th by Sens. Mark Warner, a Virginia Democrat, and Ted Budd, a North Carolina Republican. OpenAI's backing was already on the public record: the senators' announcement named it as a supporter and quoted Chan Park, OpenAI's head of U.S. and Canada policy and partnerships. Dwyer's comments, delivered during the New York City Council's livestream, reiterated that position and added her statement that the company intends to publish its analysis of AI's effects on jobs.

The bill would create a voluntary channel for AI developers and users to submit aggregated workforce data to the Department of Labor. The proposed disclosures include task-level AI use, the geographic distribution of uses and users, and changes in usage over time. The Labor Department would work with the Bureau of Labor Statistics and Census Bureau on collecting and publishing aggregated data. The bill's text and status are available through the Government Publishing Office; its latest listed action is referral to the Senate Committee on Health, Education, Labor, and Pensions.

Voluntary participation is central to the proposal. Companies would choose whether to submit information, with data de-identified and aggregated to protect privacy. That could provide a structured view of how participating organizations use AI, while leaving the resulting picture dependent on which organizations take part. The legislation describes a data-collection framework; it does not require companies to report layoffs attributed to AI or establish that a change in AI use caused a change in employment.

Dwyer's statement also points to a growing tension in how the technology's labor effects are measured. OpenAI is both a developer of widely used AI tools and a producer of research about their economic effects. The company's own studies can offer unusually direct evidence about how users interact with those tools, but usage patterns alone do not establish what happens to hiring, wages, or job counts.

OpenAI has already published research based on its products. In July, its Economic Research team said an analysis of more than 800,000 messages from U.S. ChatGPT users found that 16.8% of work-related messages and 43.5% of occupation-specific messages concerned tasks associated with another occupation. Those are measures of message content and task crossover, not a count of jobs created or eliminated. OpenAI's research archive includes that work and other studies of AI adoption and economic effects.

Dwyer leads OpenAI's central policy planning team, according to her OpenAI Forum profile. Her academic training spans astronomy and physics at Yale, aeronautics and astronautics at Stanford, and technology, management, and policy at MIT, where her doctoral research examined cost growth in complex government acquisition programs, according to MIT's profile. The policy role puts her at the junction of workforce questions and a company whose products are themselves the subject of the proposed data collection.

OpenAI has also set up an Economic Research Exchange to support outside researchers studying AI's effects on workers, firms, and institutions. The company says those projects may use privacy-protected data from its tools under defined governance and review processes. That program and the Senate bill address different routes to evidence: OpenAI's exchange is a company-run research collaboration, while the bill would place a voluntary reporting process within federal agencies.

For policymakers, the distinction is consequential. OpenAI's analysis can describe patterns in its own products, while the bill aims to assemble comparable submissions across participating AI developers and users. Neither route, by itself, guarantees a complete account of work's changing conditions. The measure's proposed disclosures are voluntary, and Dwyer did not identify a figure for job impacts during the hearing. The Senate bill remains at the committee-referral stage, while OpenAI says it will continue publishing its research.