San Jose city employees warned ChatGPT use is a public record

As widely popular artificial intelligence platforms such as ChatGPT continue to creep into our lives, San Jose is one of the first major American cities to warn its employees about the reach of the new tool, releasing a host of strict guidelines for AI use.

The city is making one thing clear up front: Be careful what you do with this new-fangled gizmo.

Everything city workers search for and create within ChatGPT is subject to the California Public Records Act, city officials warn multiple times throughout the newly released 23-page document, meaning that any citizen can get a look at how San Jose employees are using the new AI program.

“Presume anything you submit could end up on the front page of a newspaper,” the July 20 memo states on its first page, adding, “Do not use any prompts that may include information not meant for public release.”

According to city officials, some employees already are using ChatGPT.

“We’ve (heard) folks have experimented with (AI) to assist with policies, job postings, and memo sections,” wrote Deputy City Manager Rob Lloyd in an emailed response to questions about the memo. “(There are) no documented examples, yet.”

The breakthrough technology has exploded into the mainstream in recent months, raising questions and concerns about the future of work and learning.

ChatGPT, known as “generative artificial intelligence,” is already being used to write school essays, worrying educators that the tool could dramatically alter how children learn to write and form thoughts. Others are worried that AI could destroy jobs that mainly revolve around writing, such as paralegals and journalists. ChatGPT was created by OpenAI, a San Francisco-based company, and was widely released to the public in November.

Both Boston and Seattle have published similar guidelines when it comes to AI use for its workers, though only the latter considered the implications for public records requests, briefly mentioning it in a memo released in April. Public records requests, or PRAs, are routinely used by journalists, attorneys and others to gain insight into the inner workings of government, though some documents remain off limits, such as those under attorney-client privilege. Rules vary state by state.

David Snyder, who leads the First Amendment Coalition, a nonprofit that advocates for open government, said he was pleased to see San Jose bringing up public records law in its AI memo.

“I think it is good the city is acknowledging that,” said Snyder. So far, he hasn’t come across any situations where AI use by city or state government has been requested through public records, though he expects it to ramp up.

“This is a big deal,” he said. “The public has a real interest in understanding how local and state government agencies are using ChatGPT or other AI technology. The basic question is, are they using it? And how?”

Michael Morisy, the co-founder of MuckRock, a website that assists in public records requests, said he’s seen a few users ask government agencies for AI-related documents. But he confirmed no one has yet to receive the actual ChatGPT logs, which are similar to a text message interface.

The city, however, appears to be hedging on whether its employees’ AI use will always be subject to open records law.

“The technology and laws might change around what is PRA-able when it comes to interactions with Generative AI,” said Albert Gehami, who is helping oversee the rollout of AI’s use within the city, in an emailed statement. “Currently, we presume all interactions with Generative AI are PRA-able.”