The EU AI Act When considering applications of Generative AI, it is important to consider the European Union AI Act, which is also provided as a high-level summary.
Of particular importance is that the AI Act classifies AI according to its risk in four different categories:
Unacceptable risk (which is prohibited) High-risk systems (which is strongly regulated) Limited risk systems (still subject to e.g. transparency obligations) Minimal risk (which is unregulated) With some exceptions, AI systems that determine admission to education, evaluate learning outcomes or monitor prohibited student behaviour under tests are all classified as high-risk systems and need to fulfill a list of requirements that include human oversight, proper design and quality management, risk assessment, data governance and more. High-risk AI systems must undergo rigorous testing, maintain comprehensive documentation, and prioritize user safety and transparency.
...