Guiding Safe and Responsible AI Use
Oklahoma Standard AI
Definitions
Artificial Intelligence (AI) – machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or generate content influencing real or virtual environments. AI systems use machine and human-based inputs to perceive real and virtual environments, abstract such perceptions into models through analysis in an automated manner and use model inference to formulate options for information or action.
AI model – component of an information system that implements AI technology and uses computation, statistical or machine-learning techniques to produce outputs from a given set of inputs, including those such as large language model(s) or other data processing that processes information and uses computation as a whole or as a part of a system to make or execute a decision, facilitate human decision-making or can be used to communicate with clients or prospects in an automated manner.
AI system – any data system, software, hardware, application, tool or utility that operates in whole or in part using AI. All AI systems requested, developed, procured or deployed by state agencies, including standalone or embedded within existing solutions, shall be subject to the authority of the office of CIO.
Federally Protected Data - information subject to federal laws, regulations, or guidance requiring specific safeguards, including but not limited to data protected under the Health Insurance Portability and Accountability Act (“HIPAA”), the Family Educational Rights and Privacy Act (“FERPA”), Federal Tax Information (“FTI”) and Criminal Justice Information (“CJIS”).
Sensitive Data - non-public information that, if accessed, disclosed or altered without authorization, could reasonably result in harm to an individual, organization or the State, including but not limited to PII, PHI, financial information, credentials, and confidential business information.
Types of AI included in the standard
Deep Learning (DL) – refers to neural networks with many layers that learn to perform tasks by analyzing large amounts of data and is particularly strong in tasks such as image recognition, voice synthesis and autonomous systems. Examples include, but are not limited to: PeopleSoft Digital Assistance, Axon Fusus facial recognition and Adobe Sensei AI.
Generative AI (Gen AI) – models new content (e.g., text, images) that did not exist before based on training data and user inputs. Examples include, but are not limited to: Amazon Transcribe, Murf AI, Legal Files, Microsoft Viva Insights and ServiceNow Now Assist.
Hybrid AI systems – combination of multiple models used within a system that can provide a much larger range of requests by users. Examples include but are not limited to: Microsoft 365 CoPilot GCC, Azure Open AI and Hyland Software OnBase. Large
Language Models (LLM) – deep learning models for natural language understanding and generation, often trained on vast amounts of text data. Examples include, but are not limited to: Workday Assistant, Open AI ChatGPT Enterprise, Perplexity Enterprise Pro, Ask Microsoft, Ironclad Jurist, Thomson Reuters CoCounsel and Grammarly.
Machine Learning (ML) – system that improve over time through data analysis without being explicitly programmed. Often used for predictive tasks, data analysis and pattern recognition. Examples include, but are not limited to: GitHub Copilot, Otter.ai, Glean AI, Qualtrics XM, Darktrace ActiveAI and Juniper Mist AI.
Complete AI Awareness Training
The State of Oklahoma is committed to responsible, safe, and proactive use of artificial intelligence to enhance government efficiency.
