Ethics is an important part of all new technologies and is especially crucial as more people start using AI in their work and daily lives. There are risks from internal use of AI as well as from external actors who use AI to influence. In this text, we will cover some of the major ethical risks associated with AI technologies and how to think about avoiding them.
5 Risks and How to Avoid Them
1. Bias and discrimination
Automated systems based on AI technology can produce unforeseen results if one does not consider the data they are trained on and how they are used in practice. One example is training an AI to review CVs based on previous hirings. The model may then accidentally learn and reproduce the discrimination that has historically existed in society at large. If certain groups have previously had difficulty getting jobs, the AI might interpret this as them being poorer candidates. These results can also stem from an incorrect problem formulation and other non-training-related factors.
It may also be the case that a model works well and is balanced in one group, but produces discriminatory results when used in another group. This can be because the data does not represent everyone equally well or because the model has not been tested in multiple contexts.
How can this be avoided?
To counter bias, it is necessary to use representative and controlled training data and to test the system for different groups before it is used in a live setting. There should also be human review of decisions in sensitive situations, such as recruitment or healthcare.
2. AI assistants in the workplace: data protection and reliability
More and more people are using generative AI in their daily lives with the help of tools such as Microsoft Copilot, ChatGPT, and Gemini. This increases the risks of information leaks and incorrect information in their work. Since many of these tools run in the cloud and in data centers that are not always located within the EU, there is a risk that sensitive data, such as personal data (GDPR), ends up in the wrong hands or is used in ways that are out of one’s control.
There is also a risk of so-called hallucinations, where the assistant confidently provides incorrect information. This information could then end up in a final result, leading to misinformation, poor decisions, or bad recommendations.
How can this be avoided?
One way to mitigate risks is to have clear guidelines for what may be entered into AI tools, especially concerning customer data, personal data, or business-critical information. One should also always fact-check AI answers and view AI as a support, not as an authoritative source. Organizations can also choose tools with better data protection, such as AI services with clear agreements and EU storage.
3. Transparency
Another ethical risk is a lack of transparency. Many AI models function as “black boxes,” meaning it is difficult to understand exactly why the model gives a certain answer or makes a specific decision. This becomes particularly problematic when AI is used to influence important decisions concerning people, such as in healthcare, government decisions, or employment.
If one cannot explain how a decision was made, it also becomes difficult to detect errors, correct problems, and give people the opportunity to question the result. This can create a situation where AI is given too much trust simply because it feels technical and “objective,” even though it may have major flaws in practice.
How can this be avoided?
To counteract this, demands must be placed on “explainability” and documentation. AI systems used in important contexts should be able to justify their decisions in an understandable way, and there should be routines for review. Furthermore, there should always be a clear responsible person or function to handle questions and complaints about the system.
4. Disinformation and scams
AI makes it significantly easier to create disinformation. With generative models, one can produce text, images, audio, and video that look credible. Deepfakes are a clear example, where fake videos or audio recordings of famous or private individuals can be created. This can be used to influence public opinion, create conflicts, or damage people’s reputation.
In addition to political influence, AI is also used in scams and fraud. For example, a scammer can use AI to write convincing emails or texts or clone voices to sound like a boss or family member, thereby tricking people out of money or information. When AI can create content quickly and at a large scale, it becomes harder to determine what is true.
How can this be avoided?
To counter this, education and source criticism are required. People need to get better at reviewing information, double-checking sources, and being skeptical of content that feels unusual or emotionally pressing. Technical solutions with labeling of AI-generated content, tracking of origin, and better security routines (e.g., verification via multiple channels) are also important.
5. The question of responsibility
A major risk when delegating more decisions to AI systems and creating automated systems with the technology is that responsibility becomes unclear. There are various situations where this line becomes blurred. For example, a tool that can create images, a recommendation tool that suggests what kind of care to seek based on symptoms, or a self-driving car. All of these have situations where it can become unclear who truly bears the responsibility.
With image generation, offensive or illegal images can be produced, recommendation tools may give harmful advice, and a self-driving car can cause an accident. In all these cases, it is not clear whether the responsibility falls on the user, the company providing the technology, or those who created the underlying tool.
How can this be avoided?
To reduce the risks, one must have clear chains of responsibility and regulations. It needs to be clear who is responsible for what, especially when AI is used in situations where people can be harmed. There should also be requirements for testing, security, and follow-up, and AI systems should not make decisions entirely without human oversight in high-risk areas. In the development of automated systems, automation should happen gradually to guarantee safety.
Don’t forget the opportunities!
It is easy to focus on the risks of AI and forget the great opportunities the technology also provides. If ethics are taken seriously and one actively works to minimize risks, AI can contribute to much benefit. AI can automate repetitive tasks, save time, and reduce human strain, for example through automatic inspections in industry. It can also analyze large amounts of data and detect anomalies, which can help flag risks in time in areas such as healthcare and security. In summary, with clear regulations and human oversight, AI can be used responsibly and create great advantages.
More about AI
At the very core of our work is our passion for sharing and our constant desire to learn and develop. At Softhouse, we don’t just adapt to tech shifts—we shape them. By testing tools, listening to our developers, and sharing real findings, we’re building a future where AI and human expertise work together, every day. AI can feel overwhelming. It doesn’t have to be. Download our 5-minute AI guide and let us guide you.
