Obligations of Entities Under the AI Act

Per the AI Act, any AI systems that pose significant risks to the health, safety, or fundamental rights of individuals or are covered by the EU’s harmonization regulation are considered to be “high-risk”. The list and criteria of “high-risk” AI systems are dynamic, meaning they can change based on emerging technologies and their uses. 

In the case of high-risk AI systems, organizations must conduct a thorough conformity assessment that proves the system’s compliance with the relevant requirements. Such an assessment will include mechanisms to ensure data quality, transparency, human oversight, and appropriate degrees of traceability. 

Non-compliance or violations can result in fines of up to 6% of the global turnover of the organizations accused of non-compliance or violation. The fines are scaled based on the severity and implications of the violation, similar to the penalties under the GDPR.

No, AI systems developed and deployed primarily for military, defense, and national security purposes do not fall under the AI Act’s jurisdictions as it primarily focuses on the civilian uses of AI applications within a strictly business context.