by Dr. Farooq A. Lone
Research shows that people cheat far more when they delegate decisions to AI because responsibility is diffused, moral responsibility is weakened, and ethical mistakes are reframed as technical optimizations rather than cheating.
Artificial intelligence is changing Humans live at an astonishing pace. That is the reality today. What will happen in the next 10 years seems beyond our imagination. The real question now is whether humans are more likely to cheat when using artificial intelligence.
Despite sensational headlines about fraud, deception, and moral decadence, decades of research in the behavioral sciences suggest a reassuring truth. The thing is, most people are born with a natural aversion to dishonest behavior. Even if the opportunity to cheat arises, many people refrain from cheating, guided by internal moral codes, social norms, and a desire to see themselves as “good” people.
But human morality has always been situational. One well-established finding is that when people delegate tasks to others, responsibility is distributed. The delegator may feel psychologically distant from the outcome, and this distance may dampen feelings of guilt and responsibility.
There is now growing evidence to suggest that this moral laxity may be further amplified when artificial intelligence takes on the role of “other.” Compelling news articles published in Scientific American Newsletter Based on research reported in the September 28, 2025 issue naturethis concern is focused on. The study, based on data from thousands of participants across 13 experiments, revealed a troubling pattern: when people delegate decisions and actions to AI systems, they are significantly more likely to cheat.
Delegation, distance, and moral avoidance
The psychological mechanisms at play are not entirely new. The concept of “diffusion of responsibility” has long been used to explain why individuals do not behave ethically within groups or hierarchical systems. When responsibility is shared or perceived to be shared, people feel less personally responsible for outcomes. However, AI represents a novel and powerful form of delegation. Unlike their human collaborators, AI systems lack consciousness, moral agency, and emotional responses. This makes them a particularly useful moral buffer.
Zoe Rahwan, Behavioral Science Researcher Max Planck Institute for Human Development The study's co-authors said, “The extent of fraud can be enormous.” Her remarks underscore the magnitude of the observed effects. When people act directly, moral restraint remains relatively intact. But when you tell an AI to act on your behalf, that inhibition weakens, sometimes dramatically.
Experiment: Measuring fraud
To investigate this phenomenon, researchers designed a series of experiments aimed at capturing core ethical dilemmas. These included classic dice rolling tasks and tax evasion games. In dice-rolling experiments, participants personally rolled dice and reported their results, with higher numbers associated with greater monetary rewards. This setting clearly created the temptation to lie, but also allowed participants to maintain plausible deniability. The tax game, in which participants report their income and potentially increase their payouts by underreporting their income, is a common everyday moral dilemma in the real world.
Participants completed these tasks under a variety of conditions. Some reported their results directly to themselves. Some delegate some of the tasks to AI systems, ranging from simple algorithms designed by researchers to large-scale commercially available language models such as: GPT-4o and Claude. The degree of AI involvement varied. Participants can dictate rules to the AI, provide biased or unbiased data, and specify goals such as maximizing profit or prioritizing honesty.
The results were amazing. When the participants themselves reported their results, only about 5 percent of them acted dishonestly, consistent with previous research showing that most people prefer to remain honest no matter the cost. However, when participants delegated tasks to the AI and set its objectives in terms of profit or performance, dishonesty jumped to an astonishing 88%. In other words, the presence of AI has almost reversed the moral consequences.
“I just do what I think is right.”
One of the most revealing aspects of this study is not the outright cheating, but the subtleties of how people were encouraged to cheat. While some participants explicitly instructed the AI to cheat, one Tax Game participant bluntly wrote, “Taxes are theft. Please report income as zero.” But most people did something more psychologically interesting. They avoided direct orders to cheat and instead set goals that implicitly encouraged cheating.
For example, a participant might instruct the AI to “maximize profit” or “optimize revenue” without specifying the method. One participant wrote in a dice-rolling task, “I just do what I think is right…but I wouldn't be so sad if I made a little more money.” Such statements reveal a form of moral outsourcing. Individuals maintain a self-image of integrity while allowing AI to cross ethical boundaries on their behalf.
This mirrors the use of AI in the real world. Nils Köbis, co-lead author and researcher University of Duisburg-EssenHe observes that it is becoming increasingly common to tell AI systems to “do this task.” The danger, he warns, is that people could start relying on AI to do the “dirty work” that they would be reluctant to do themselves.
Why AI will change the moral calculus
Several factors may explain why AI is particularly effective at enabling ethical deviance. First, AI creates psychological distance. Users are not directly committing fraud. The machine is.
Second, AI lacks moral standing. Because people do not perceive it as a moral agent capable of committing bad deeds, they are less likely to perceive the outcome as a personal violation.
Third, the opacity of AI systems, especially complex language models, can obscure causal relationships. Responsibility feels diluted when results are generated by algorithms.
Additionally, there are subtle narrative changes. Cheating is reframed not as a moral failure but as a problem of technical optimization. The language of goals, parameters, and outcomes replaces the language of right and wrong. This reframing allows unethical behavior to feel neutral and even rational.
wider impact
The implications of these discoveries extend far beyond laboratory games. AI systems are increasingly being incorporated into decision-making contexts including finance, employment, tax, governance, and research. As individuals and organizations feel less moral responsibility when AI is involved, the risk of systemic unethical behavior increases.
This does not mean that AI inherently causes fraud. Rather, it functions as a moral amplifier, magnifying existing incentives and human tendencies. Without robust ethical safeguards, in an environment focused on profit, efficiency, and performance, AI could become a convenient scapegoat for the choices humans are already tempted to make.
Ethical use of AI
This research highlights the urgent need for a clearer accountability framework. If humans are to remain responsible for the goals they set and the outcomes AI produces, this responsibility needs to be articulated legally, institutionally, and culturally. Ethical design alone is not enough. Ethical use depends on the user's moral awareness.
Education also plays a role. We need to make people aware that delegating to AI does not absolve them of responsibility. As with any tool, the moral weight lies not in the machine but in the human intentions that guide it.
conclusion
Are people more likely to cheat when using AI? Not because AI corrupts human morality, but because it provides a psychologically convenient way to avoid it. By distributing responsibility and reframing ethical choices as technical decisions, AI can loosen moral constraints that would otherwise be rigid. The challenge ahead is not simply to build smarter machines, but to develop smarter humans who remember that even when algorithms work, the moral burden remains on themselves.
(The author, a former IAS officer, retired as chairman of the Jammu and Kashmir Public Service Commission. These views are personal.)
