By Doug Levy

As local governments and other organizations increasingly adopt artificial intelligence (AI) technologies, the need for responsible implementation becomes paramount. AI integration presents opportunities and challenges, particularly in ensuring that these powerful tools are used ethically and transparently. Here, we explore ways that organizations can develop policies that reflect their community's or their organization’s values and enhance administrative efficiency, audience engagement and strategic communications while also protecting privacy and data security and avoiding legal or other risks.

Both employees and external stakeholders or constituents may be apprehensive about AI, especially if they have not used it. A well-crafted policy goes a long way toward providing everyone with comfort about AI and its immediate implications.

“Every city needs an AI policy to keep up with the massive tech shifts that seem to happen every 10-15 years—like cloud computing 15 years ago, mobile before that, social networks, and even the rise of the internet and the advent of desktop computing,” says David Oro, a member of the American Canyon, California city council.

“AI is the next big leap, and whether we like it or not, we’ve got to adapt with smart policies that help us move forward in a safe and productive way.” - David Oro, council member, City of American Canyon.

Elements to include in AI policies

AI policies must establish a framework for using AI but do not need to be exhaustive. Key elements are: 

1. When and whether AI may be used

2. Who must review or approve specific AI tools before they are used

3. Whether AI-generated content requires disclosures or other labels

4. How to use AI within data security and privacy rules

5. When or whether any data or other content may be uploaded to AI

6. Who must review AI-generated content before it is released

7. How new AI tools may be considered for implementation

8. Who will review or update the policy and on what schedule 

AI is more than an IT issue

While AI policies must begin with a consideration of your information technology (IT) infrastructure, they must reflect much more than IT policy. Thus, the appropriate owner of AI policies is most likely a chief administrative officer, whose decisions can reflect input from IT, communications, legal, and other sectors. One of the most important reasons for an AI policy is to alert employees about unrecognized risks. For example, using an AI note-taker in a Zoom or other meeting has the potential to break attorney-client privilege or violate privacy rules.  

Some agencies identify specific AI tools within their policies, while others leave those details out but designate where such information is compiled. Choose whichever is more appropriate for your organization. Just be sure there is a clear method for considering new tools, reviewing existing ones as they update or change,  and information about tools that are not allowed. For example, some agencies have a brief policy document and a side memorandum that lists authorized tools. Some also include a list of AI systems that were considered but rejected, along with a short explanation. 

For public agencies, transparency is especially important, and government agencies must communicate clearly about how AI is or will be used within their operations. This involves detailing the specific applications of AI, what data is used, and how the output is used. and explaining the underlying algorithms and data sets that inform these systems. For instance, if a city employs an AI tool to assess public safety risks based on crime data, it should disclose what data is being analyzed and how the results influence decision-making. Transparency fosters trust among constituents who may be wary of technology's implications on their lives. Explaining how AI assists professionals to do their work faster -- without fundamentally changing the work itself - may be one step in providing needed reassurance. 

The second critical aspect is accountability. Local officials must recognize that with the power of AI comes responsibility for its outcomes. Establishing clear lines of authority includes designating proper oversight to ensure adherence to ethical principles, avoidance of bias and compliance with privacy and other relevant rules.  For local governments, an AI oversight committee also can provide a useful platform for citizen input.

Moreover, engaging stakeholders throughout this process is important both to improve compliance and build or maintain trust. Customers, residents, all stakeholders should have a voice and be heard. Community engagement can take many forms: public forums where citizens share their thoughts on proposed uses for AI technologies, surveys gauging perceptions about privacy concerns, or collaborative workshops where residents work alongside policymakers to plan or review AI implementation. Local governments can create an inclusive environment that respects diverse perspectives by actively involving constituents in discussions about AI use (or other sensitive issues.)

Protect against bias or other AI-driven mistakes

Another significant issue surrounding responsible implementation relates to algorithmic bias—which has gained increasing attention as more organizations rely on automated systems to assist decision-making. Bias occurs when algorithms produce skewed results due to flawed training data or inherent prejudices embedded within their datasets. For example, if you ask a public AI tool to create a narrative, check whether the characters created in the story accurately reflect the situation. In many instances, AI will generate stories with characters based on stereotypes – such as women as nurses or teachers, men as mayors or governors, etc. 

This kind of bias, based on the training data used to develop the AI systems, raises serious ethical questions and jeopardizes accuracy.  Restricting AI use to tools that are vetted and programmed with your own data sources is one way to protect against this. Implementing AI requires use of diverse data that accurately reflects a community’s demographics and other factors. 

Furthermore, government agencies would benefit from regular audits of their AI use by internal or external organizations. These assessments would help identify any unforeseen biases within deployed algorithms and provide insights into improving performance and more equitable decision-making. 

“A policy is essential to define parameters and provide permission for using AI in an organization while also serving as a framework for ensuring safe and responsible use.” - Alexa Davis, assistant city manager, Rolling Hills Estates, California. 

Implementing AI also requires more than distributing a written policy, especially given the apprehension that many individuals have about AI. To facilitate all aspects mentioned above—transparency, accountability, stakeholder engagement—the development process for adopting new technologies should also include training tailored toward equipping staff with knowledge of ethical considerations and the nuts-and-bolts operation of generative AI within the organization and its workflows. 

“Whether your organization is public or private, the time to adopt an AI policy is NOW. ” says Brian Baker, founder/CEO of Big Sky Crisis Communications. 

"Your policy doesn’t need to be perfect and it doesn’t need to cover every new AI technology. Rather, it should be flexible enough to accommodate new technologies and new ways of using AI to work smarter and more efficiently." - Brian Baker, crisis communications expert.

Training initiatives could cover topics such as understanding best practices for including AI tools in content generation, data analysis and basic research. Workshops are great places to compare the results of a query about your community made on a public AI tool with a tool within your organization's firewall so that staff members recognize the difference and why only one is acceptable for professional use. 

There are many useful resources for those interested in developing AI policies. If you are interested in guidance or customized workshops on implementing AI, we can help.

In addition to our consulting and training services, here are a few references that may help: 

• Carmichael, Mary. "Key Considerations for Developing Organizational Generative Ai Policies." @ISACA, Updated 1 November 2023, 2023, accessed 12-20-24, 2024, https://www.isaca.org/resources/news-and-trends/newsletters/atisaca/2023/volume-44/key-considerations-for-developing-organizational-generative-ai-policies.

• County of Sonoma. "Sonoma County Adopts Artificial Intelligence Policy." 9/10/2024. https://sonomacounty.ca.gov/sonoma-county-adopts-artificial-intelligence-policy.

• Rolling Hills Estates, California, Generative Artificial Intelligence Policy, adopted September 24, 2024.

• TrustCloud Corp. "Artificial Intelligence Usage Policy Template.” accessed 12-20-2024, https://community.trustcloud.ai/docs/trustops/helpful-resources/documentation-templates/artificial-intelligence-usage-policy-template/.

• Workable Technology Limited. "Key Considerations for Developing Organizational Generative Ai Policies." In AI tool usage policy, Workable Technology Ltd., 2024. https://resources.workable.com/ai-tool-usage-policy.

A version of this article was published by iQ 360® at https://iq360inc.com/blog/corporate-public-policy/responsible-ai/.

© 2024 Doug Levy Communications LLC. All rights reserved. May not be reproduced or copied in any form without written permission.