What the “AI is just a tool” argument gets wrong

This way of framing technology has become almost automatic in public debate. It appears constantly in discussions around artificial intelligence, biotechnology, and many other emerging technologies. The underlying idea is always roughly the same: technologies themselves are neutral, while responsibility lies primarily in the hands of users.

At first sight, the statement sounds reasonable. It reflects a deeply rooted way of thinking about technical objects. A knife can be used to prepare food or to injure someone. The object itself is not moral or immoral; what matters is the intention of the person using it. Applied to AI, the conclusion appears straightforward: the technology is not the problem. The challenge is simply to encourage good uses and prevent harmful ones.

But I increasingly think this analogy fails to describe what contemporary technological systems actually are.

The problem is not that technologies can produce both positive and negative effects depending on how they are used. Of course they can. What this analogy misses is that many modern technologies do far more than passively wait for individual users to decide what to do with them. They actively shape which behaviors become easier, more profitable, more legitimate, and eventually impossible to avoid.

From technological choice to technological inevitability

Seeing technology as neutral often appears alongside another common fallacy: AI is already here, so the only remaining question is how to make its uses ethical, responsible, or safe.

Both framings quietly remove something essential from the discussion: the possibility of questioning the technological trajectory itself. Public debates often move quickly from the question of whether certain technological trajectories should exist at all to the question of how they should be governed responsibly. The development and deployment of AI systems are treated as largely inevitable. The remaining discussion concerns risk mitigation, ethics frameworks, safety protocols, transparency standards, or responsible use guidelines.

Yet this framing already contains an important political assumption: that the trajectory itself is no longer really open to collective choice. The issue is no longer “Should we develop and deploy these systems in this way?” but rather “Now that these systems are here, how do we mitigate their effects?” In other words, governance becomes focused on adaptation rather than on the possibility of collectively shaping or limiting technological directions upstream.

Why technologies are not “just tools”

This matters because technologies are never deployed into a social vacuum. The idea that technology is neutral or that only "mitigation after the fact" is up for debate fails to describe how our technological systems actually emerge.

Let’s go back to our classic example: a knife can be used to cook or to kill. The object itself is supposedly neutral; morality depends entirely on the user. But large-scale modern technologies do not operate like knives.

A knife does not come with global financial incentives pushing toward specific uses. Entire industries are not organized around maximizing knife adoption. Governments do not redesign public infrastructures to integrate their use. Universities do not restructure research agendas and priorities to optimize them. Massive lobbying ecosystems are not built to normalize knife dependency across society.

Artificial intelligence, however, exists within exactly these kinds of dynamics.

Its development is shaped by enormous concentrations of capital, geopolitical competition, state strategies, industrial interests, military contracts, and research budgets. The uses that become dominant are not random. They emerge from systems of incentives and infrastructures that actively orient how the technology is designed, deployed, and integrated into everyday life.

This is why describing AI as “just a tool” is misleading. Not because individual responsibility disappears, but because individual intentions alone are insufficient to understand how technological systems evolve. A researcher may genuinely want to use AI to accelerate scientific discovery. A hospital may deploy AI systems hoping to improve diagnostics. A public administration may adopt automated systems to increase efficiency. None of these actors necessarily intend harm.

And yet the broader consequences may still include increased energy consumption, concentration of power among a handful of technology companies, new dependencies on opaque infrastructures, intensified surveillance capacities, or transformations of work that nobody fully controls. The issue is therefore not simply the morality of individual users. It is the structure of the socio-technical system itself.

The limits of the pharmakon

French philosopher Bernard Stiegler was among the thinkers who most powerfully challenged the idea of technological neutrality. Drawing on the Greek concept of the pharmakon, he argued that technologies are always simultaneously poison and remedy.[1] Writing, for example, can preserve knowledge but also weaken memory. Digital technologies can connect individuals while also producing attention capture and alienation. For Stiegler, technology could never be reduced to a neutral instrument because it always transformed human ways of living and thinking.

This remains an essential insight. But contemporary technological systems raise an additional problem that goes beyond the question of ambivalence alone.

The issue is not simply that technologies contain both danger and potential. The issue is that technological systems increasingly organize the conditions under which some uses become dominant, profitable, normalized, and scalable. They do not merely allow certain behaviors; they actively orient collective trajectories.

Social media platforms optimized for engagement are not neutral communication tools occasionally misused. Their architecture systematically incentivizes particular behaviors because attention extraction is built into their economic logic. Similarly, artificial intelligence systems developed within global competition dynamics, capitalist accumulation, and state surveillance infrastructures cannot be understood independently from those environments. The systems surrounding emerging technologies during their development and deployment make some outcomes vastly more probable than others.

Importantly, none of this requires bad intentions. Many actors involved in these transformations are acting rationally within the constraints and incentives they face. Institutions seek productivity gains. Governments seek strategic advantage. Companies seek markets. Researchers seek funding opportunities and publication capacity. Each individual decision can appear reasonable in isolation while still contributing to broader dynamics that nobody explicitly chose democratically.

Re-politicizing technological systems

This is precisely why the distinction between a tool and a socio-technical system matters politically. When technologies are framed as neutral tools, responsibility is individualized. Harm is explained primarily through misuse, unethical actors, or insufficient regulation. But once technologies are understood as systems structured by incentives, dependencies, institutional behavior, and social expectations, responsibility can no longer be reduced to individual morality alone.

The question then becomes much broader. It concerns how technological trajectories are collectively organized long before end users make any explicit choice. It concerns who finances technological development, which applications become economically viable, which forms of expertise shape deployment, and which visions of society become embedded into technical infrastructures.

This is also why current debates around “responsible AI” sometimes feel strangely insufficient. Very often, responsibility frameworks focus on reducing harms while leaving the overall direction of technological development largely unquestioned. The possibility that societies might collectively choose different trajectories, slower deployments, or even refusals in certain domains rarely enters the discussion in any meaningful way.

Yet this is fundamentally a political question, not merely a technical one.

What is ultimately at stake in the “AI is just a tool” argument is not merely a semantic disagreement. It is a political question about agency. If technologies are treated as neutral and inevitable, then collective deliberation becomes secondary. Society is reduced to adapting to trajectories already defined elsewhere.

But if technologies are understood as socio-technical systems shaped by institutions, incentives, infrastructures, and power relations, then technological development becomes something that can, and should, remain open to democratic questioning.

Not every technological trajectory needs to be accelerated simply because it becomes technically feasible. Not every application deserves deployment because it is profitable. And not every social transformation should be accepted as the unavoidable consequence of innovation.

The real challenge is therefore not only to use technologies responsibly once they exist. It is to recover our collective capacity to decide which technological systems we actually want to build in the first place.

[1] Bernard Stiegler develops the concept of technology as pharmakon across several works, notably What Makes Life Worth Living: On Pharmacology (Polity, 2013), drawing on the original Greek meaning of pharmakon as both remedy and poison.