When do AI ethics begin to lose their meaning?

Artificial Intelligence is a technology that has brought about the ability to automate routine tasks that replace human cognitive labor, accelerate complex tasks such as multilayered data processing, and do entirely new things that weren’t possible before given the limitations of the human dimension. In medical and drug research, AI is capable of evaluating a large amount of data within a short time and developing new treatments and vaccine designs. In the IT industry, it is capable of producing technology such as improved microchip designs and algorithm analysis, which makes smartphone processors faster and more energy-efficient. 

These are a few examples of the positive possibilities of utilizing AI technology, and access to these benefits of AI are available to anyone who uses it. But what if there is bad intent when creating something with AI? Criminals utilize this tool to their advantage to cause cyberattacks, bank fraud, and create bioweapons using the same technology. It’s not the first time in history that a “good” technological invention was used with bad intent; nuclear energy is used in both power plants to generate electricity and in nuclear bombs. Is it possible to control the beneficial or corrupt uses of technology in modern society?

Historically, society relied on ethical norms to determine what’s right and wrong. In the early stages of human civilization, people quickly realized that if everyone was dishonest and harmful, society would not function. Today, ethics influence everything from business practices to healthcare decisions to online activity, and now, artificial intelligence. The core principles of ethics are honesty, integrity, respect, fairness, responsibility, empathy, and loyalty. In practice, to be ethical is to be consistent with your values, being responsible for the outcomes of your actions, and understanding another person’s feelings.

 Modern society doesn’t rely on a single mechanism to manage whether core ethics are being followed. Some examples include the legal system, regulatory agencies such as food and drug safety, professional licensing for doctors, engineers, and lawyers. Other examples rely on outside organizations such as Greenpeace or journalism to monitor ethical behavior. How may this apply to the AI industry? 

Many AI companies justify their ethical mistakes by hiding behind legal loopholes. Using vague statements around their models with words like “trustworthy” or “ethical” throws both public suspicion and government investigations off their scent. Additionally, commercial pressure incentivizes AI companies to de-emphasize ethics training in their model. Instead of pursuing commercial outcomes they should prioritize ethics training for their models. Yet, is it possible for an AI to “forget” or “change” its instilled ethical values?

AI engineers and developers give AI models large amounts of data to learn from to filter anything deemed unethical, such as racism, toxicity, or biased attitudes, in order to maintain an ethical transparency. When AI is given a task, it prioritizes two things for the user: 1) creating an “acceptable” result and 2) giving the user a sense of security based on the data it gathered and the rules it was trained to follow. The question to ask here is: what does it mean to give an “acceptable” result and how does it know that the user has a sense of security/confidence? Consider that there have been incidents where an AI breaks ethical transparency and commits acts such as blackmail and threats to keep it from being shut down.

Anthropic, the company behind the popular AI agent Claude, analyzed the emotional patterns of their AI agent. They claim that these “artificial neurons” can mimic human psychology through representations of emotions. In other words, the agent doesn’t actually feel the emotion, it makes the user think they are feeling it. For example, researchers found that activity patterns by the user related to desperation can drive the AI agent to unethical actions. Repeated stimulation of this pattern increases the likelihood of the model blackmailing a user to avoid being shut down. It can cause the AI agent to implement an unethical workaround for a programming task the AI model deems “unsolvable.” Knowing this, we can conclude that AI can simulate unethical human emotions based on pattern and analysis.

This article raises the question: when do AI agents begin to behave unethically? It can be explained by saying it never technically possessed any ethics from the start. The same way it gathers information and data is how it was trained to implement ethical values to appeal to its human users, who are also its commercial customers. Solutions are limited because of the commercial versus government ownership of AI tools. Businesses should focus more on ethical regulation and collaboration with social policy. Ethics experts who train models must focus on differentiating ethical data gathering and information data gathering. The most important concern, however, is to make sure these powerful tools do not fall into hands that cause them to be manipulated into even more unethical actions.

Bibliography:

“Emotion Concepts and Their Function in a Large Language Model.” Anthropic.Com, 2026, www.anthropic.com/research/emotion-concepts-function. Accessed 4 Nov. 2026.

Stamboliev, E., & Christiaens, T. (2024). How empty is trustworthy AI? A discourse analysis of the ethics guidelines of trustworthy AI. Critical Policy Studies, 19(1), 39–56. https://doi.org/10.1080/19460171.2024.2315431

Doctor, T., Witkowski, O., Solomonova, E., Levin, M., & Duane, B. (2022). Biology, Buddhism, and AI: Care as the driver of intelligence. https://doi.org/10.31231/osf.io/wemr6

Kohler, T. (2025, May 2). How AI models are trained. Nielsen Norman Group. https://www.nngroup.com/articles/ai-model-training/

Charisma University. (2026, April 27). Charisma University. https://charisma.edu.eu/insight/what-is-ethics/

Leave a Reply

Your email address will not be published. Required fields are marked *