The Tangwall Campagin. At work, at the doctor’s office, in our search for information or our travels… Artificial intelligence is already making its way into most areas of our lives. In your workyou emphasize that the main issue around AI is not to oppose or not its development, but to ensure that we used it with confidence. What do you mean by that?
Tiffany Xingyu Wang. The debate for or against is already outdated. The AI is no longer knocking on the door, it is already in the house. My children, who are one and three years old, talk to Alexa and ride in self-driving cars with me in San Francisco like it’s commonplace… For them, it’s not the future. It’s the decor.
In this context, there are two ways to fail. The first: placing excessive trust in AI. This amplifies misinformation, automates bias and lets deepfakes destroy reputations in a matter of hours. Our intuition is no longer enough: the signals we have always used to know who to trust (confidence, ease, familiarity) are precisely those that AI imitates best. The other pitfall would be to show excessive distrust of AI: we would then leave medical advances and climate solutions blocked in laboratories. Only calibrated confidence allows us to learn and progress. This is the whole purpose of my book: to make trust in AI a discipline that we practice, and no longer a reflex.
One of the specificities of this technological revolution is that it takes place, precisely, in a context of generalized distrust – towards those in power, the media, scientific authorities… “ AI arrives in a world that is not capable of welcoming it »warn yourself. How to overcome this crisis of confidence?
If we wanted to imagine the worst moment for the arrival of AI, we would choose this one! Ten years ago, we were talking about post-truth. We have now entered the post-trust era. Post-truth was emotion prevailing over facts. With post-confidence, it’s more serious: we have lost the common rules that allowed us to establish what is true. Until now, “seeing was believing”. The photo of the Pope in a white down jacket, generated by AI and believed by millions, marked the end of this era.
The way out of this crisis of confidence will be through reconstruction, not persuasion. We do not decree trust, we build it. We must check when designing AI (what sources are they based on?), keep humans in the loop for serious decisions and clearly designate someone responsible when something goes wrong.
You write, about inventions that have disrupted human societies such as the railway or electricity: “We spent decades building the conditions that made them trustworthy. » AI systems are being deployed “faster than we can build guardrails”. Should we hope for a pause in the development of AI?
A general pause is neither realistic nor necessarily desirable. Three engines simultaneously power AI development : the scientific quest, the race for profits (more than 400 billion dollars were invested by companies in 2025) and geopolitical fear. No actor alone can stop this machine. The company that takes a break to refine its models risks being left behind for good…
But “inevitable” does not mean “uncontrollable”. The inevitability of AI relies on a system of incentives; however, these incentives can be modified. Resisting does not mean stopping progress. It’s about building the muscles that progress tends to skip: assessment, accountability, alignment. This is what makes the speed bearable.
“It’s not the AI that needs to be paused. It’s our reflex of blind trust”
Concerning the risks linked to AI that we do not yet understand, such as emergent behaviors or the strategic deception of certain models, the rule is simple: when we do not know how to govern what we build, we must admit it and stop. The real break is individual: I invite everyone to pause before delegating a decision to a machine. This pause depends neither on Washington nor on Beijing. It depends on us.
In Europe, the fear of having fallen irremediably behind in the AI race in relation to the United States and China continues to be raised. Among American and Chinese professionals in the sector, is the old continent already perceived as offside ?
We often hear from Silicon Valley that Europe regulates while others innovate. Yet in the United States, the “we can’t get left behind China” argument is too often used to justify faster deployment and lighter safeguards. This is not a model to copy with your eyes closed.
Europe has a strategic asset: it knows how to set standards. The GDPR has become a global reference. The AI Act bases its approach on human dignity and fundamental rights. In a post-trust world, this is a competitive advantage, provided you move from principles to execution. In this regard, the Paris Summit marked a turning point: in 2025, the emphasis shifted from defining principles to implementation and industrialization, with significant investment commitments. The challenge now is to move quickly without giving up what makes Europe strong.
“Industrial Europe has its place in the AI race”
Singapore is an interesting example in this sense: the city-state has made trust an infrastructure, through testing, auditing and certification tools that companies actually use. The result is less fear and more experimentation. In Germany, Siemens Energy is using AI to power electricity grids and wind turbines in the North Sea. An invisible AI, which has gained trust because it works. Industrial Europe has its place in this AI.
Let’s return to the individual level: to be able to trust an AI on a daily basis, you propose in your book a real method to follow. If it had to be summarized in a few sentences, what does it consist of?
I called my method “CALIBRATE”. This is not a checklist, but nine questions to ask yourself before relying on AI:
- Contextualize : where and for what is AI used?
- Align : Who does this system work for – mine, or the platform, the advertisers…?
- Restrict : at what point should humans take control?
- Integrity : what is this system based on, how is its knowledge updated, and how errors are detected and corrected?
- Bias : Do I trust her only because she confirms what I think?
- Claim guarantees of reliability : does AI work in real conditions, and not just in demonstration?
- Assign responsibility : if things go wrong, who is responsible?
- Transparency : where do the answers provided by this AI come from?
- Exercise judgment : am I still thinking, or just validating?
If in a hurry, the short version of this method is “LIT”: Limit, Inspect, Trace. In other words: know the limits of an AI, question its foundations and go back to the source.
Among the risks posed by AI, you mention that we gradually lose our ability to carry out certain tasks without it. Do you think that will future generations still know how to write, drive, cook… reason, without the help of AI?
The erosion has already started, gently. GPS has made us worse navigators, autocorrect has weakened our spelling, recommendation algorithms have narrowed our findings. Each loss seemed small. AI will accelerate this phenomenon on the scale of a civilization. The figures are there: according to a global study by KPMG, 64% of employees put in less effort because they know they can count on AI, and 45% let it do its thing rather than learning. An MIT study shows that after a few months of assisted writing, most users struggle to explain a text that they have just produced themselves!
“The real risk is not the replacement of humans by AI, but deskilling and dependence”
The real risk is therefore not the replacement of humans by machines, but deskilling and dependence. The answer is to voluntarily keep a part of the effort that maintains critical thinking, and reinvest the time saved in what the machine cannot do.
The question to ask yourself every day: does this tool make me more capable, or more dependent? As a mother, I don’t want my children to reject AI. I want them to remain in control of their relationship with her, and not to have it decided for them.
* Graduated in law from Fudan University, holds a master’s degree in international finance from Sciences Po Paris and an MBA from the Wharton School, Tiffany Xingyu Wang is the founder of the AI laboratory Songsheet. She has been distinguished by the American magazine Forbes as one of the “Most Entrepreneurial CMOs” of the year 2022.
** The Trust Code: How to Unlock AI’s Promise Without Losing Humanity (“The code of trust: unlocking the potential of AI without losing our humanity”), John Wiley & Sons, not yet translated into French.
