On this page...
Purpose
This Policy aims to promote responsible and ethical use of generative artificial intelligence (AI) technologies. By establishing clear minimum requirements and prohibited uses of AI, we ensure that AI technology is utilized to enhance efficiency, innovation, and productivity while safeguarding against potential risks and misuses. All Support Entity workforce members (employees and contractors) and any contracted third-party performing work on behalf of the Agency must comply with this Policy. Supported Entity senior leadership is responsible for establishing procedures for their Agency’s compliance with the requirements of this Policy. This policy is binding on Supported Entities and should be considered a best practice for others.
Back to top
Background
Generative AI technology, including chatbots, virtual assistants, and similar applications, is increasingly prevalent in professional settings. These technologies, such as OpenAI's ChatGPT and DALL-E, Google Gemini, and Microsoft Copilot, offer diverse functionalities and integration options, ranging from standalone systems to seamless embedding within existing infrastructures. While generative AI tools have the potential to significantly enhance productivity, efficiency, and innovation across various tasks, including drafting documents, text editing, idea and image generation, software code writing, data analysis, and anomaly detection, they also pose risks such as bias, inaccuracies, and intellectual property concerns. As AI technology evolves rapidly, organizations must carefully assess both the benefits and risks associated with its adoption. The State of Iowa is actively developing procedures to navigate this dynamic landscape of AI technology.
Back to topReferences
Back to topPolicy
When using generative AI tools and capabilities, this policy applies to all business use cases, including but not limited to:
● content generation including audio, images, written documentation (i.e., policy, legislation, contracts), and correspondence (i.e., letters, memorandums, text messages, emails);
● development of software code;
● research;
● presentations and training;
● summarizing and proofreading documents; and
● decision making.
Supported Entities and their workforce members are accountable for the decisions that are made and materials created using generative AI.
AI Use Requirements
Software Code Generation
When considering the use of AI generated code, Supported Entities must consult their legal counsel to determine whether AI generated code should be allowed.
In cases where computer code is generated, written, or modified using AI technologies, Supported Entities must thoroughly identify and mitigate all business and security risks related to its use.
If AI is used to write a portion of code, that particular section of the code must be “walled off” by a comment at the start of the section noting that the code was created with AI, and a comment at the end noting where the AI stopped. AI generated code cannot be protected with intellectual property rights.
AI Solution Failure
If AI tools do not meet the needs of the business or service, human fallback systems must be in place. This includes human intervention if AI solutions fail. These alternatives must ensure continuity of services.
For example, a chatbot powered by AI may serve as the initial point of contact for citizens seeking government services. Citizens should have the choice to exit the AI tool, enabling them to engage with a human through the chat window or receive contact information for communication with a human representative. Citizens should always be given the option to leave the AI tool so that they can interact with a human.
Training and Awareness
Training must be provided on how to use AI ethically and responsibly, including lessons on privacy and security. Ongoing awareness efforts must be provided to keep workforce members informed on the changing world of AI and the best ways to protect privacy and security. All AI training curriculum identified for use by Supported Entities must be approved by the Department of Management before it is made available.
Back to topProhibited Uses of AI
It is strictly prohibited to use AI technologies for activities that may pose ethical concerns, are harmful, illegal, or in violation of state policy. Prohibited uses of AI include, but are not limited to:
Use of Sensitive and Protected Data
Sensitive (i.e., internal communications) and protected data must not be used to train an AI tool, submitted into an AI system, or utilized by an AI tool (i.e., automated meeting note taker, audio recording and real time transcription) without prior written authorization by the Department of Management.
Blind Trust of AI Outputs
Prior to finalizing any output generated by AI technologies, the Supported Entity workforce member must subject such output to thorough human evaluation and iteration that is appropriate for the task at hand. This review should focus on checking if the AI-generated content is accurate, relevant, and suitable. Responses generated must not:
● be used verbatim;
● be treated as the sole source of reference; or
● be the sole source of making final business decisions.
Personal Accounts
Personal accounts utilizing AI tools and applications cannot be used for work-related purposes.
Manipulation of information
Do not use AI to create or spread fake or misleading information to individuals.
Intellectual Property Rights and Infringement
Ensure that AI used for content generation (e.g., artwork, music, software, etc.) does not infringe upon intellectual property law.
If a body of work needs to be protected by copyright or some other intellectual property protection, users of generative AI tools must recognize that the output of such tools cannot be protected by intellectual property rights. Moreover, the inclusion of AI outputs in a larger body of work may render the entire work as incapable of being protected by intellectual property rights. Supported Entities and their workforce members must consult their legal counsel whenever intellectual property protections are implicated.