A future-proof AI policy starts by recognising that technology can support humans, but that human oversight remains crucial. Especially in contexts where AI has a direct impact on human lives or fundamental rights, a hybrid approach is indispensable. Limitations such as hallucinations, bias, embedded biases, and a lack of transparency about decision-making processes illustrate the need for human intervention. Full factual reliability of AI cannot be guaranteed today.

Therefore, human involvement is not an optional layer, but a fundamental starting point in the design and application of AI: from the outset, it should be clear where technology stops and where humans make decisions about the purpose, boundaries and control points. In domains such as recruitment and selection, medical diagnosis or credit assessment, explicit human intervention is not a luxury, but an ethical requirement.

AI systems should not only be smart, but also reliable. This means they must be resistant to errors, external influences, and abuse. Robustness goes beyond avoiding bugs: it’s about resilience to unexpected inputs, security against tampering, and control over responses to disturbances.

A reliable AI system recognises its own limits and, if necessary, can be temporarily stopped or adjusted, either automatically or under human supervision. This creates a controlled, safe interaction between humans and machines, even in emergency situations.

Organisations that incorporate this properly not only strengthen their technological reliability, but also the trust of employees, customers, and society.

Privacy is not a side effect, but a core value. It is linked to fundamental rights such as autonomy and the right to a private life. Privacy in AI means more than protecting personal data from abuse or data breaches. It requires structural respect for the control that individuals should retain over their data, even when that data is processed through complex algorithms.

Embed privacy in the design phase, not afterwards: with clear agreements, technical safeguards and transparent processes.

Transparency is fundamental for responsible use of algorithmic systems. It is about clarity on how an AI system works, what data it uses, and what logic it applies to make decisions. Transparency does not mean that everything need to be publicly available. It is about targeted, context-specific communication: who needs what information, in what form, and for what purpose?

For employees, this means staying critical, asking questions, and actively contributing to clear communication about how AI works in their organisation. Transparency is not an end goal, but an ongoing process that contributes to responsible and credible technology.

At all times, AI systems should ensure that no person is disadvantaged due to gender, ethnicity, age, disability, socioeconomic status, or other personal characteristics. While this principle seems obvious, in practice we still often see that this is not always the case. When training data are not representative, minorities are often not represented or are represented incorrectly. This can lead to models that categorise, exclude, or favour people without any explicit malicious intent behind it.

AI should not be a goal in itself, but it should contribute to the well-being of people and society. Being aware of this is important because the technology carries an ecological and social footprint, ranging from high energy consumption to reinforcing inequality and increasing digital exclusion. Only by consciously choosing social added value over technological hype does AI contribute to society.

Organisations that take this wider impact seriously build better technology, but also a society where technology serves people.

Whoever uses AI in an organisation bears responsibility for the operation as well as the consequences of the system. Without clarity, risks might shift without anyone really being accountable. It is therefore essential that organisations invest in solid governance frameworks today. Not as a precaution for future legislation, but as a strategic choice to avoid risks, reputation damage, and social resistance.

Outsourcing responsibility to technology is an illusion. At a time when AI influences or even makes decisions, it must remain crystal clear who is morally and legally accountable for it. Only then will technology retain its legitimacy in a just society.