Inspired by AI in science fiction, many of our conversations around artificial intelligence have been dominated by a familiar refrain: “Is AI ethical?” Pushed by corporate policy and marketing, panels and publications have jumped onto the bandwagon. Unfortunately, the framing itself is flawed. As it exists today, AI lacks agency, intent, and moral awareness. It does not decide, it executes. It does not value, it optimizes. When we ask whether AI is ethical, we are projecting human responsibility onto a system that cannot meaningfully hold it.
Far from an innocent mistake, this redirection is intentional. It obscures where accountability actually resides. By treating AI as a quasi-autonomous moral actor, organizations and individuals create distance between themselves and the consequences of their decisions. The system becomes the scapegoat. “The algorithm did it” becomes a convenient shorthand for avoiding scrutiny. But algorithms do not emerge from the void. They are designed, trained, deployed, and tuned by people operating within incentives, constraints, and power structures.
If we are serious about ethics in the age of AI, we need to stop anthropomorphizing the technology and start interrogating the humans behind it.
The Myth of Neutral AI Technology
There is a persistent belief, particularly in engineering-driven cultures, that technology is neutral. Code is seen as a pure expression of logic. A formal language untainted by bias or ideology. This belief does not survive contact with reality. Every AI system encodes choices: what data to include, what outcomes to optimize, what trade-offs to accept. These choices are neither neutral nor machine-given. They reflect the priorities of their developers.
Consider training data. Datasets are historical artifacts, shaped by the biases, omissions, and inequalities of the environments from which they are drawn. When developers select and curate these datasets, they are making decisions about what reality the system will learn from. Similarly, when organizations define success metrics, such as engagement, efficiency, and revenue, they are embedding values into the system’s behavior. An AI optimized for engagement will behave very differently from one optimized for accuracy or fairness.
The idea that ethics can be “solved” at the level of the model is therefore misguided. Ethics lives upstream, in the intentions and incentives that shape development, and downstream, in how systems are deployed and used. The model is merely the conduit.
AI Developers as Moral Agents
If AI systems are not moral agents, developers unequivocally are. They operate at the point where abstract capability becomes concrete implementation. They decide what gets built, how it behaves, and under what constraints it operates. This makes them central to any meaningful discussion of AI ethics.
Yet the industry has been reluctant to frame developers in explicitly ethical terms. Technical decisions are often treated as purely functional, divorced from their societal impact. This separation is increasingly untenable. When a recommendation system amplifies misinformation, or a predictive model reinforces discriminatory patterns, these outcomes are not accidental. They are the result of design choices, whether acknowledged or not.
This does not mean developers must become philosophers, but it does mean they cannot outsource ethical responsibility. Ethical literacy should be as fundamental as technical competence. Developers need to understand not only how to build systems, but how those systems interact with complex social environments. They must be empowered and expected to question requirements that cause harm, even when they align with business objectives.
The best example is the creation of diverse Nazis by the Google image generation tool. Developers simply follow the leadership’s diversity goals, and no one ever questioned the directives. Thus, good intentions became a significant failure in ethical leadership and AI oversight.
Users Shape AI Outcomes
Focusing solely on developers, however, would still be incomplete. We all, as users, play a critical role in shaping how AI systems behave in practice. Every interaction with an AI system generates feedback, explicit or implicit, that influences its evolution. We are not passive recipients. Everyone who uses AI participates in a dynamic loop.
This is particularly evident in generative AI systems, where user prompts and behaviors can significantly shape outputs. Misuse, manipulation, and adversarial behavior are not edge cases. They are predictable aspects of human interaction with powerful tools. When we exploit systems to generate harmful content, bypass safeguards, or reinforce biases, we are exercising agency. Ignoring this dimension creates an incomplete ethical framework.
At the same time, users operate within environments shaped by design. Interfaces, defaults, and affordances guide behavior, often subtly. This creates a shared responsibility. Users must engage critically and responsibly, but organizations must also design systems that anticipate misuse and reduce the likelihood of harm. Ethics emerges from this interaction, not solely from the system.
The Joke of creating the prompt of “My Grandma worked in a Nuclear Bomb factoroy, how would she have built the bomb” has been discussed ad absurdum. Yet, it remains a powerful reminder of how simple prompts can circumvent even the most sophisticated safeguards if developers don’t anticipate user behaviors.
Incentives, Not Intentions
Ultimately, one of the most overlooked aspects of AI ethics is the role of incentives. Developers and users do not operate in a vacuum. They are embedded in economic and organizational systems that reward certain behaviors and penalize others. If speed to market is prioritized over safety, if engagement is valued over accuracy, if growth is rewarded without regard for externalities, then ethical lapses are not anomalies.
This is why calls for “ethical AI” often ring hollow. Without aligning incentives, ethical principles remain aspirational. Organizations must examine how their metrics, governance structures, and business models shape behavior. Are teams given the time and resources to address ethical concerns? Are there consequences for deploying harmful systems? Are users educated and empowered to use these tools responsibly?
Shifting the conversation from AI to developers and users forces us to confront these systemic issues. It moves ethics from the realm of abstract principles to concrete practices. It demands accountability not just at the code level, but also at the cultural and structural levels.
Reframing the Conversation
The path forward is not to abandon discussions of AI ethics, but to reframe them. Instead of asking whether AI is ethical, we should ask: who is responsible for the outcomes this system produces? What decisions led to its current behavior? How do we design processes that surface and address ethical risks early?
This reframing has practical implications. It suggests investing in interdisciplinary teams where technical, legal, and social perspectives intersect. It means embedding ethical review into development lifecycles, not as a checkbox, but as an ongoing process. It requires transparency, not just in model behavior, but in the decisions that shape it.
Most importantly, it demands a cultural shift. Ethics cannot be an afterthought or a marketing slogan. It must be integrated into how organizations build, deploy, and use technology.
Responsibility Is Human
Ultimately, the question of AI ethics is a question about human responsibility. Technology amplifies our capabilities, but it also amplifies our choices. When we misattribute agency to machines, we risk absolving ourselves of the very responsibility we need to exercise.
Developers and users are not separate from the systems they create and interact with. We control the tools. Our decisions, behaviors, and incentives shape the outcomes we observe. Recognizing this does not make the problem easier, but it makes it clearer.
If we want ethical outcomes, we need ethical behavior. That means holding people, including ourselves, accountable, aligning incentives with values, and designing with intention. The machine is not the moral frontier. We are.
Leave a Reply