Employment contracts in the age of AI: are staff quietly signing away their employment identity?

UK employees may be giving up far more than they realise when they sign “standard” employment contracts in an AI-driven workplace. Such contracts must evolve quickly to meet this challenge.

For years, intellectual property (IP) provisions in employment contracts have contained broad IP assignment provisions in the employer’s favour. The aim was simple: ensure that anything created “in the course of employment” belonged to the employer. But what happens to personal IP that employees bring with them?

Knowledge workers generate a constant stream of IP: reports, emails, policies, software code, playbooks, and datasets. Increasingly, employers can aggregate and even monetise that output by feeding it into AI systems, creating derivative tools replicating an individual’s style and reasoning.

Emails sent over years of employment effectively give AI the ability to simulate an employee long after they leave.

Comparisons with the film industry are hard to ignore. When an actor signs a contract they assign or license rights in their performance, and the studio can exploit that performance indefinitely. Actors now commonly negotiate “AI carve-outs” to protect themselves. These clauses explicitly prohibit the use of performances to train AI systems for new content.

However, an employee who signs an employment contract may, without realising it, be permitting their employer to store, mine and adapt their “image” by training their AI tools to preserve and use that employee’s professional “voice” as a reusable and monetised asset.

The “voice”, which a senior employee develops over their career, is not usually carved out and protected in their employment contract (unlike pre-existing client relationships which are excluded from the scope of any termination restrictions in the new contract).

Employers will rely on standard IP clauses drafted for a pre-AI world to argue such use is already permitted. Very few currently mention AI explicitly.

The legal risk for employers lies less in ownership and more in consent, transparency and scope. Employees may argue they did not give informed consent to their work being used in ways that dramatically extend its lifespan and reach.

Overlaying this is the emerging regulatory framework around data protection and AI, including the EU AI Act and anticipated UK guidance, which is expected to constrain how employee-related data and IP can be used to train AI models.

Future employment contract provisions will likely spell out the extent to which employee-generated IP can be used for AI training, what happens on termination, how pre-existing personal IP is protected and whether limits apply to internal or external exploitation of the employee “image”.

Everyday senior UK employees sign away their professional “brand”, failing to appreciate that IP waiver clauses operate much more like film deals – encompassing the value of not only everything they create whilst they are employed, but also employment IP they bring with them.

Photo by Aleksandr Kadykov