The development of artificial intelligence creates new prospects in areas that include cybersecurity and raises critical ethical questions. When using AI in vital applications like safeguarding personal information or managing energy structures, it is crucial to assess the impact of wrong algorithms. Some examples include intrusion detection systems and intrusion prevention systems that employ machine learning models to monitor and analyze data in real-time to detect threats. The main issues are the system's opacity (the so-called ‘black box’ issue), the possibility of bias in the data used, and legal liabilities. The lack of transparency can lead to difficulties in understanding AI decisions, thus affecting user trust and allowing attackers to exploit algorithms. Bias in data can lead to unfair decisions and also tends to weaken security mechanisms. Also, it is still unclear who will be responsible for the AI failures, particularly regarding the data breaches and financial losses. To address these challenges, it is essential to promote the development of Explainable AI technologies, teach people how to use AI, check data, and adopt new legal requirements for AI. The paper analyses these problems and calls for a cross-disciplinary strategy encompassing technological, ethical, and legal perspectives to make the use of AI in cybersecurity more reliable.