What primary aspect does safety in AI systems aim to minimize?

Safety in AI systems primarily aims to minimize AI harms, such as misinformation and disinformation. This focus arises from the increasing integration of AI in various aspects of society where incorrect or harmful outputs can have serious consequences, such as influencing public opinion, spreading false information, or making biased decisions.

By prioritizing the minimization of these types of harms, AI safety protocols are designed not only to ensure the reliability and accuracy of the AI outputs but also to safeguard ethical standards and societal norms. This is especially important in applications that can impact public health, security, and trust in information sources.

Other options, while important in their own right, do not align with the central aim of safety concerning the harmful impacts of AI on individuals and society. Data redundancy, energy consumption, and model complexity are relevant to AI performance and efficiency but do not encapsulate the core purpose of safety measures, which is to protect users and the public from adverse effects stemming from AI technology.