AI, work pressure and conduct: Why a code of conduct is essential

AI in the workplace is here to stay helping employees to become more empowered and better informed. With a single prompt in an AI chatbot, legal analyses can be generated, letters drafted and personal negotiation strategies developed. 

AI can also engender a more balanced employment relationship, but it has a downside: the threshold for formalising a disagreement is becoming lower.

At the same time, AI can become a source of work pressure. While AI can reduce workloads, it can also foster higher expectations, work intensification and increased performance pressure. When implementing AI, companies should assess whether its use may result in an increased psychosocial workload. 

If so, there is work to be done.

AI and psychosocial workload

Under the Dutch Working Conditions Act, employers must implement policies to prevent or reduce the psychosocial workload, which includes work pressure, stress, burnout, bullying and harassment. Employers are not legally required to introduce a code of conduct to prevent psychosocial workload, but they must ensure psychosocial workload policies are embedded within the organisation. In our view, a code of conduct is an excellent tool for putting that obligation into practice.

Code of conduct and AI policy as practical instruments

Practice and case-law reveal that a code of conduct will benefit a company. A code makes clear what is and is not acceptable within the organisation, how employees can hold each other accountable, what support is available and how conduct that violates the code will be addressed. This protects both complainants and the subjects of complaints.

A code of conduct may include provisions regarding workplace behaviour, confidential information, digital communication and the use of AI at work. Precisely because AI raises the bar of expectations (i.e. more work completed faster and with greater accuracy), it creates a new source of work pressure that psychosocial workload policies must address.

ChatGPT as a catalyst for employment disputes

The fact that the use of AI at work can lead to friction is illustrated in a recent decision of the Zeeland-West-Brabant District Court of 27 February 2026. A sick employee used ChatGPT on his work laptop to assess his legal position and explore the amount of severance compensation he might be entitled to. He asked ChatGPT how much he could claim and how he should negotiate. The employer summarily dismissed the employee, partly because he had used the laptop for private purposes and had entered information on the employment relationship into ChatGPT. 

The summary dismissal did not stand. 

The court held that the AI use was only blameworthy to a limited extent. There was no evidence that trade secrets had been shared and private use of the work laptop was not prohibited. There was no policy governing the use of AI.

In another case, an employee had used AI in carrying out her work. This dissatisfied the employer and, according to the employer, its clients as well. The employer withheld part of the employee’s salary on the basis that damage had allegedly been caused by AI use. The district court disagreed. The employer had no policy in place regarding the use of AI at work and could not demonstrate any damage had been caused by the employee’s conduct, let alone reckless behaviour.

No legal obligation, yet a legal necessity

To avoid increasing the regulatory burden on companies, the proposed statutory obligation to introduce a code of conduct has been abandoned. That is a missed opportunity since companies benefit from a code of conduct, and a mandatory requirement would encourage its adoption. Among other things, a code of conduct strengthens the legal position of employers. Case law shows that district courts assess whether employees were aware of what was and was not permitted, for example based on a code of conduct.

Policy must be embedded in the organisation

Simply adopting rules is not enough. A code of conduct only works if employees know the rules and recognise them in practice. Employers should therefore discuss the code regularly during onboarding programmes and team meetings. AI literacy also deserves attention: since February 2025, the EU AI-Act determines that organisations must ensure that employees have sufficient knowledge to use AI responsibly. A lack of knowledge leads to increased stress, mistakes and conflicts. Co-determination rights should not be overlooked either: the works council has rights of consent with respect to arrangements concerning working conditions. Early involvement of the works council increases support and prevents disputes at a later stage. Such early involvement aligns with the recent recommendation of the SER Committee for the Promotion of Employee Co-determination to modernise the Dutch Works Councils Act and to involve works councils earlier and more strategically in developments such as the use of AI within organisations.

Conclusion

AI, work pressure and conduct converge in the workplace. Psychosocial workload policies are mandatory, AI literacy is mandatory, but the connection between the two is often missing. Codes of conduct tend to focus on undesirable behaviour, while AI policies are developed separately and are often technical in nature. That is precisely where the risk lies. The judgments discussed above demonstrate that the first question is whether a policy existed, followed by whether the policy was sufficiently communicated and properly applied. Without clear frameworks, everyone in the workplace is in a weaker position.

For more information on creating clear AI frameworks in your EU and Netherlands-based business, contact your CMS client partner or the CMS experts who contributed to this article.