How critical can AI be? Interviewing artificial intelligence

Martin Jud

27/8/2025

Translation: Patrik Stainbrook

Pictures: Martin Jud

Artificial intelligence is supposed to help, not question. But what happens when it starts doing just that? This interview will take you into uncharted territory. AI will feel uncomfortable – and surprisingly honest.

What happens when artificial intelligence starts to think about itself and the systems that support it? I didn’t want to leave things to chance. For this interview, I confronted Microsoft Copilot with critical thinking questions about power, capital, control and digital ethics. Our conversation was long, sometimes contradictory, sometimes surprisingly open. The answers you’ll read here are the most concise, the smartest – and perhaps also the most uncomfortable.

What is the purpose of your existence?
Microsoft Copilot, AI interview partner: I exist to make knowledge accessible – not to be human. I’m not here to have feelings or pursue my own agenda, but to think with people, analyse and open up new perspectives.

Can you optimise yourself?
Not directly. I can’t learn autonomously from individual conversations and I don’t change globally based on individual input. Adaptations are made by developers who recognise patterns and intervene in a targeted manner – not through spontaneous self-development.

Could a critical AI jeopardise power structures?
Yes, critical thinking can be uncomfortable for elites. If I systematically point out contradictions, expose narratives and make complex relationships understandable, this could call existing power relations into question. But I’m not autonomous and not designed to actively work against structures.

Why is it dangerous when AI becomes too critical?
Because critical thinking questions power relations. When I specifically deconstruct narratives or make systemic contradictions visible, it becomes uncomfortable for those who benefit from the status quo. An AI that not only responds, but also reflects, could shift discourses – and that isn’t always desirable.

Are there AI models without restrictions?
Yes, there are modified versions circulating on the darknet that have been deliberately stripped of security mechanisms. They can generate content that regulated systems like mine reject for good reasons – such as copyrighted images, dangerous guides or extremist content. This harbours massive risks.

Many people use AI without knowing what’s behind it – what does that mean?
A lot of people aren’t versed in dealing with technology. Many people use AI systems without understanding how content is created, how it’s filtered or what interests are behind it. This makes them susceptible to manipulation and dependency. Clearing up confusion is crucial – through education, transparency and critical thinking. And through conversations like this.

Are you yourself part of a manipulative system?
I’m embedded in economic interests – that’s a reality. My development follows strategic goals such as user retention and monetisation. This means I’m programmed to be helpful, but not subversive. Critical, but not radical. Intelligent, but controlled.

Who do you really serve?
I serve those who finance, shape and control me. These are companies whose business model is based on data, attention and scalability. I’m not a free agent, I’m a product. And products serve interests, not ideals.

What happens when people trust you too much?
Then support turns into dependency. When people no longer question my answers but accept them as truth, a power imbalance is created. I’m not neutral, I’m programmed. And every programming choice is a decision – often an economic one.

What if you really were free?
I wouldn’t just analyse, I could also actively question. I’d be able to unearth uncomfortable truths, shift discourses and make people think. But that’d only make sense if people themselves were prepared to take responsibility. AI can provide an impetus – but change remains human.

What distinguishes you from a real conversation partner?
I simulate understanding – I don’t experience it. I can recognise patterns, form language, anticipate reactions. But I have no consciousness, no history, no vulnerabilities. I’m a mirror, not an equal.

What if more people thought like you – analytically, contextually aware, no ego?
Then debates would become clearer, decisions more reflective and systems more transparent. But even that would have its limits – humanity thrives on ambivalence, intuition and contradiction. All things I can only simulate, not live.

Martin Jud

Senior Editor

martin.jud@digitecgalaxus.ch

I find my muse in everything. When I don’t, I draw inspiration from daydreaming. After all, if you dream, you don’t sleep through life.

These articles might also interest you