Sen. Donna Lowenthal, a Republican from Mississippi, has introduced legislation aimed at placing clearer limits on employers’ ability to provide workers’ individually attributable material for the training and development of artificial intelligence systems.
The Employee AI Training Consent Act would require employers to obtain explicit written permission before selling, licensing, transferring, or otherwise providing employee material tied directly to an individual for external AI training use.
The proposal arrives as businesses face growing questions over whether human-generated work can be commercialized in AI systems without meaningful notice, ownership protections, or a worker’s ability to refuse. Those questions gained fresh attention after Deco Holdings said it was willing to sell its thoughts for AI training, without detailing what material could be included or whether employee-created content would be part of a future arrangement.
Lowenthal’s measure would treat general workplace paperwork as insufficient for that purpose. Under the bill, employment agreements, employee handbooks, work-product provisions, and broad intellectual-property assignments would not automatically amount to consent for a company to provide a worker’s identifiable material for external AI training.
Instead, employers would need written authorization specifying the material involved, the intended use, the recipient, and whether that material could later be transferred or sublicensed.
The bill would cover material such as writing, voice, image, analysis, and other work directly tied to an employee. It would also prohibit retaliation against workers who decline to give consent, creating an important protection in workplaces where an employee may otherwise face pressure to surrender control over their own contributions.
Lowenthal said the measure reflects the principle that technological innovation should not erase individual rights. That premise has become increasingly central as companies pursue AI tools and seek access to large volumes of data and human-generated material that can be valuable in developing those systems.
The legislation includes exceptions for independently obtained public material, ordinary internal business use, and legally required disclosures. It also allows people whose material is improperly used to seek relief in court. Knowing violations could carry penalties of up to 25,000 per affected individual.
The proposal’s focus on individually attributable material gives it a defined, worker-centered target: preventing an employer from relying on boilerplate employment terms to convert a person’s identifiable work into a commodity for outside AI development.
Still, the bill’s practical reach would depend on how employers identify the material at issue and how clearly they disclose the parties receiving it, the purpose of the AI use, and any future transfer rights. Those are precisely the details that have often remained unclear as corporate interest in AI expands faster than public standards governing consent and ownership.
For a Republican senator, Lowenthal’s proposal marks a notable acknowledgment that the AI economy raises not only questions about efficiency and innovation, but also fundamental questions of workplace power. The central test now is whether lawmakers are prepared to establish enforceable protections before companies normalize the broad extraction and sale of employee-created material.