When a top city official said Fresno was being held “captive” by private companies and their proprietary software, she meant it.
The surprisingly candid conversation took place publicly during a budget hearing last month, when Councilman Miguel Arias began searching for answers about how much the use of artificial intelligence is costing the city.
“I suspect AI will be like meth,” Arias said at the June 8 budget hearing. “They’ll give it to you for a really cheap price the first day, and then the second year they’ll rob you and rape you because we’re on the hook now.”
Moments later, Fresno City Manager Georgeanne White explained how the City of Fresno is held captive by private companies that develop software, programs and technologies critical to the city’s core functions.
“They don’t have these standalone products that you just upload anymore — we’re stuck,” White said in a later interview with Fresnoland. “We’re stuck because if you don’t keep your product up to date, they stop maintaining it, they stop updating it. We’re stuck in a way and it just becomes the default.”
White remembers a time in the mid-1990s when, after using WordPerfect for so long, she didn’t want to switch to Microsoft Word – which was new at the time and quickly gaining popularity. She must have been the last person at Fresno City Hall to hold on to the older word processor, White told Fresnoland, until she could no longer use it while everyone else had already moved on.
She sees the current emergence and use of artificial intelligence in a similar way. With advances in technology, it feels like there isn’t much choice, she said.
“Either you’re going to have technology or you’re not,” White said. “We’re going to need technology and that’s why we’re trapped – we’re absolutely trapped.”
This dilemma doesn’t just affect the city of Fresno. It highlights the broader societal impact that the emergence of artificial intelligence has had, as well as the potential impact on the future of work or the environmental impact of AI data centers on water, energy and even the cost of phone and computer parts such as RAM.
The inevitability of AI — or at least the spirit of it — has reached Fresno local government. City employees have been using AI products for professional purposes for years. Fresno City Hall has no formal policies setting guardrails for how AI can and cannot be used.
Additionally, city leaders—particularly most members of the Fresno City Council—seem unaware of how much deploying artificial intelligence in the city actually costs. Even fewer have an understanding of how AI impacts the way people think and solve problems – something that is currently understudied by researchers.
So many city of Fresno employees were using OpenAI’s ChatGPT that the company contacted city officials and asked them to pay for an enterprise subscription or stop using it altogether, White told Fresnoland.
At that point, Fresno City Hall blocked the use of ChatGPT — specifically its domain address on city servers — and directed staff to instead use CoPilot, which is integrated with Microsoft Office, as an alternative.
Although Fresno City Hall does not have a formal policy on the use of AI, White emphasized that an informal policy already exists and it is simply a matter of formalizing it in an administrative order.
While discussing the necessary guardrails, White brought up a recent legal embarrassment at an outside law firm representing the city of Fresno.
“You see what happens in the court system where lawyers are tripped up and sanctioned, right?” White said. “It happened to us that one of our outside lawyers used an invalid quote that probably came from AI.”
Fresnoland received a memo from the city listing two dozen different programs that city employees are allowed to use that either contain AI features or are solely AI products. The memo lists projected costs for some of them, but does not list actual costs in previous years.
A city spokesperson told Fresnoland that the actual costs of AI programs, including Claude AI tokens, are not readily available.
According to the memo, the city of Fresno plans to spend about $25,000 in the next fiscal year on Madison AI, software that creates internal employee reports. They also expect similar costs to use AI built into Salesforce as well as Wordly AI, which provides live transcription and Spanish translation during city council meetings.
But when it comes to Claude AI, which is used by the city’s information services department, the expected cost is not clear because it relies on utility tokens, which incur separate costs. City staff made no projections about how much utility tokens would cost in the next fiscal year.
Fresno police officers use AI integrated into Axon’s body cameras and software to create police reports. Although there are no separate costs associated with using these AI capabilities, the city’s $18.3 million contract with Axon shows how costs can pile up when private companies provide software and equipment that the city is tied to.
Although the first year of the city’s five-year contract with Axon costs $1.2 million, the third year of the contract will cost three times as much and the fifth year will leave the city with a much larger bill of $5.8 million.
Table of Contents
Is AI really inevitable? Or is it a choice?
As investment in AI increases, an emerging area of AI security has raised alarms and urged caution in recent years. It is made up of scientists, engineers and researchers who have formed a coalition that is pushing technology companies to address the consequences of the technology they develop before releasing it to the public.
Top OpenAI employees and executives have even resigned from their positions in protest, among other things, at the company’s stance on the responsible development of AI tools as well as its own transparency, accountability and documentation.
David Krueger, an assistant professor at the University of Montreal who studies AI, has been one of the most vocal proponents of AI safety over the past year. In 2025, he founded Evitable, a non-profit organization with the aim of preventing the uncritical development of AI programs and technologies that could have consequences far worse than those the industry has already experienced – some of which he described in the Guardian.
Krueger told Fresnoland that AI poses a threat to security and privacy, can sometimes hallucinate, is quick to agree with users in a so-called sycophancy, and spits out information based on gender, political or other biases.
That’s why it’s important to set clear guidelines for the use of AI in the workplace, said Krueger.
“It is critical for governments to have policies in place regarding the use of AI, both to protect their constituents and for transparency,” Krueger told Fresnoland, adding that ideally such policies should have come first before government employees began using AI for professional or professional purposes.
He told Fresnoland that an AI policy should include maintaining a record of all types of AI usage and exclude any AI products that might reveal private data or train data based on inputs to the system. He also said a policy must ensure employees are aware of the possibility of AI hallucinations and sycophancy.
Krueger also made a sobering statement about current AI advances: “On the current path, AI will increasingly automate decision-making at all levels and eliminate human decision-makers. This is very dangerous and could cause humanity to lose control of our destiny.”
He said it was a decision to adopt AI technologies and workflows, but it was also a decision to move away from it – something some large companies have done recently. Back in May, Uber announced that it had exhausted its AI budget for a full year by April, and its COO said that spending on AI services was becoming increasingly difficult to justify, according to Business Insider.
The CEO of the multibillion-dollar company specifically said that while Uber’s engineers are using AI more than ever before, it doesn’t translate into better features or experiences for customers.
That same month, Starbucks canceled an AI inventory tool it rolled out in thousands of stores across the country after it frequently miscounted and mislabeled items, according to Reuters.
The question of how AI could transform jobs has also focused on how the use of artificial intelligence affects how a user thinks and processes information. An article published in April 2026 by researchers at the University of Pennsylvania’s Wharton School discussed how the use of AI can “supplement or even replace” human thinking to such an extent that a user must experience what the article calls “cognitive capitulation.”
The researchers found that when AI is introduced into a given workflow, a user does not construct a response but instead adopts the response provided by the AI.
The paper referred to artificial intelligence as System 3 and concluded that “humans not only use System 3 to help them think, but are often devoted to its results – whether correct or incorrect.”
“We don’t just use AI, we think with it. In doing so, we have to ask new questions: What happens when our judgments are shaped by a mind that is not our own?” the researchers asked in the paper’s conclusion.
“What happens to intuition and effort when a generative, artificial partner is ready to respond? How do we preserve agency, reflection, and autonomy in a world where users engage in cognitive abandon?”
This article first appeared on Fresnoland and is republished here under a Creative Commons Attribution-NoDerivatives 4.0 International license.
https://fresnoland.org/2026/07/10/artificial-intelligence-software-tech-captive/
