Cities make progress with guidelines for generative AI tech

The city has established baseline guidelines for employees on the use of generative artificial intelligence technology for city business and document production, but is still determining policies regarding the use of the tools by vendors and other outside parties.

The efficiencies and expanded skill sets enabled by AI tools could lead to changes in the city’s workforce, according to a recent memo.

The memo, released late last month by Chief Innovation Officer Daniel Culotta, follows a City Council resolution approved in February that aimed to create rules and best practices around the rapidly evolving technology available to platforms like ChatGPT and Bard.

In response, the Office of Innovation convened an advisory committee made up of three City departments (the Department of Communications and Technology Management, the Department of Information Security, and the Department of Equity) as well as outside experts and stakeholders. The group’s goals were to foster collaboration inside and outside City Hall, create an evaluation process for new technologies, identify and monitor inherent risks, prioritize transparency, equity, and accountability, and find responsible ways to expand technology use.

Much of the city’s research and guidance on using generative AI is contained in a “Generative AI Standards” document that was compiled this fall and approved for use based on guidelines established by the city of San Jose, California.

The standards document covers many of the factors that public officials must consider when using generative AI tools: Because submitted queries and prompts will be publicly available for the purposes of AI training, the standards document states that any data used will automatically become a record available to public information requests.

It also mandates that output from AI tools must be carefully fact-checked and reviewed before final approval, and whenever AI tools are used to generate city documents or other deliverables, city staff are instructed to disclose their use and their origins.

The city also adopted three risk levels covering the impact of a breach of data used in AI tools: medium risk, high risk and prohibition.

Medium risk uses include internal documents containing non-sensitive information where there are no issues with identifying personal data.

High risk uses include personally identifiable information or non-public business information that could impact the City’s business processes. High risk material will not be allowed in the generative AI system.

Prohibited risky uses include sensitive identifying information, such as credit card numbers, bank account information, health data or other personal data.

The memo also references several recommendations made in recent months regarding the use of generative AI, but does not specifically state which recommendations city officials are currently using.

Staff is also determining how to address incidents of error or bias related to AI tools, as well as “considering the implementation of procurement review guidelines, vendor guidelines, fact sheets, and other tools that would promote transparency, explainability, understanding, and auditability of the structure and functionality of AI applications the City is evaluating or using.”

The Innovation Office also leads the City’s efforts in training and use of AI technologies, with more than 300 employees receiving instruction on responsible AI use.

Looking ahead, the memo states, “The Office of Innovation will work with the Deputy Chief Learning Officer to research and identify additional resources to expand AI training and education options. The AI ​​Core Team will develop a strategic plan with the Deputy Chief Learning Officer and Head of Human Resources to comprehensively address preparing, reskilling, and supporting staff as AI tools continue to redefine the workforce.”

of Austin Monitoris funded by donations from our community. We occasionally feature donors in our reporting, but we are careful to be transparent and separate our business and editorial efforts. You can see a full list of our donors here, and our code of ethics is explained here.