Gangnam Perspective: AI Ethics and the 'Free Lunch'

When ethics enters business, people tend to tense up first. That is because of the naive idea that ethics is a cost that eats into profits. The industry was also highly wary of the government's recently announced 'Artificial Intelligence (AI) Ethics Principles.' That was true even though they are voluntary guidelines with no legal force. There is, of course, a reason for that. The principles could later serve as the basis for self-assessment checklists for companies and may also influence future revisions to related laws.
But Korea's AI ethics standards are softer than they may seem. That becomes clear when looking at the higher-level law behind the AI ethics principles, the AI Basic Act. The European Union's AI law uses the term 'high-risk' to describe regulatory concerns. It defines the problems AI products or services can pose to humans as risks. By contrast, Korea's AI Basic Act uses the term 'high-impact.' The word 'impact' is neutral, neither positive nor negative. Caught between civic groups calling for stronger ethics standards and the industry demanding lighter regulation, the oddly vague term 'high-impact' is a watered-down expression. Since the parent law for the AI ethics principles is written this way, a middle ground for compromise has already been built in.
Rather than worrying about regulatory uncertainty, it may be better to change the shallow way we think about ethics. Consumers are often eager to free-ride on the benefits of AI. We also make contradictory demands of AI. We want AI to know us well and protect us conveniently, while still keeping our personal data as secure as possible. We want AI to handle tedious tasks, but we want to retain control over important decisions. We want AI to be fast and powerful, but we cannot tolerate errors or accidents. We demand that companies behind AI bear heavy responsibility, yet we do not want service launches to be delayed or prices to rise. If we discuss AI ethics while stubbornly denying that convenience comes with costs and responsibility, it will end as nothing more than armchair theory.
Companies also need a more practical shift in attitude. AI ethics is not about drawing a line between good and evil. It is a process of finding the right balance among personalization and privacy, automation and human control, the speed of innovation and the time needed for verification, and corporate efficiency and social safety. In practice, weak corporate ethics often lead to legal disputes and painful consequences. Recent leaks of personal and corporate data are cyber incidents. Discriminatory hiring or failed financial investments using AI algorithms damage a company's reputation. If an AI agent exercises the wrong authority, it can bring services or even the company's systems to a halt. AI ethics is now reaching into legal affairs, security, quality control, investment risk, reputation, and business continuity.
AI ethics should now be viewed as an equation of cost and investment. The idea that AI ethics and industrial competitiveness stand on opposite sides is outdated. Take Meta Platforms as an example. The company agreed to pay up to $18 billion over 10 years in lawsuits filed by dozens of U.S. states over social networking service addiction among teenagers and related mental health harms. It paid an astronomical price for digital ethics issues. At the same time, those ethical standards also function as a massive investment, because they create barriers that make it harder for competitors to enter the market.
AI ethics standards are destined to change more often and become stricter over time. The principles unveiled this year are an upgraded version of the 2020 standards. The reason they were revised again after six years is the enormous technological shift that occurred in the meantime with the arrival of ChatGPT. In 2020, algorithmic bias, personal data, and explainability were at the center of AI ethics. But as AI's Use of Knowledge in Society became mainstream, copyright, hallucinations, deepfakes, misinformation, and overdependence emerged as new issues, making the 2020 ethics standards insufficient.
The pace and influence of technological change will only accelerate from here. As AI agents emerge, public-sector AI transformation expands, and physical AI deployment begins in earnest, it is obvious that new ethical versions will soon face problems they cannot handle. The business opportunities lie not in avoiding the ethical debates triggered by AI, but in responding to them quickly and proactively. There is no free lunch in AI ethics.
[email protected] Cho Chang-won Reporter