GenAIGuidelinesForGovernmentENCompressed

Fairness It is crucial to implement measures to identify groups affected by the system and to prevent or minimize biases, discrimination, or profiling that individuals or groups may encounter due to data or algorithms. This is essential to avoid negative discrimination against specific groups. Generative artificial intelligence tools possess the capability to generate content that may ex- hibit discriminatory tendencies, lack representation, or contain biases and stereotypes. These biases can be diverse and linked to various intersecting identity factors, such as gender and race. A number of models are trained using vast datasets from the Internet, which is often the source of these biases. Users must: 1. Carefully evaluate the created content to ensure alignment with the Kingdom of Saudi Ara- bia’s regulations, values, and ethics. This evaluation must also include comprehensive and care- ful scrutiny of biases or stereotypical associations. 2. Learn to write commands (for GenAI tools) in ways that ensure the creation of correct and free of biased content. 3. Make efforts to obtain a comprehensive understanding of the data used to train the tool, in- cluding knowing the source of the data, its contents, and how it is selected and prepared. 4. Enhance users’ understanding and awareness of bias, the importance of diversity and inclu- sion, anti-racism, values and ethics, as this knowledge will contribute to improving their ability to identify biased content. Generative Artificial Intelligence Guidelines For Government 3.2 Reliability and Safety A generative artificial intelligence system must ensure adherence to specifications, operating consistently as intended and expected by its designers throughout all stages of its life. Reliability serves as a measure of the credibility and dependability of the system in carrying out its specific functions and achieving intended results. Generative artificial intelligence tools are often suscep- tible to inaccuracies, as they learn from internet data containing errors and outdated information. While content generated with generative artificial intelligence may initially seem consistent, it frequently lacks the logical consistency or context necessary. Recognizing both strengths and limitations is crucial in the responsible use of generative artificial intelligence within government entities. Ensuring quality necessitates a multi-faceted approach encompassing monitoring, hu- man supervision, and user education to effectively harness the capabilities of generative artificial intelligence and reduce associated risks. Users must: 1. Identify content generated with generative artificial intelligence to ensure that users are pre- pared to address potential reliability issues and have the ability to verify content using other sources. 2. Proofread content for factual accuracy and relevance to context to prevent the spread of misinformation. 3. Seek to understand the quality and sources of training data used by a generative artificial intelligence system to enhance content reliability. 4. Encourage supplementing content created with generative artificial intelligence with informa- tion from trusted sources to ensure accuracy. Generative Artificial Intelligence Guidelines For Government