Artificial intelligence lurked throughout Thursday’s meeting of the Somerville City Council, including adoption of a law setting boundaries for its use in government, protecting residents from decisions made using AI and declaring that the technology shouldn’t be used to replace employees or creative work.
The law and civil rights orders were sent to the council’s Legislative Matters Committee for a deeper look, but as discussion at a Tuesday committee meeting showed, keeping AI out of government work – or even setting the rules on what that includes – won’t be simple.
The proposed ordinance begins by saying city departments can’t use AI in law enforcement actions including arrests, searches and investigations, yet the committee meeting this week brought a report from staff that the city was using only one true AI product – and it was for police.
The software, called Blue Voice, was created by David Lawrence, a self-described Harvard Law dropout, to summarize laws, policies and procedures and advise police officers in real-life decision-making, giving them what company materials call “the right answer in seconds.” It is a “closed-loop” system, meaning it is trained only on Somerville policies and transmits no data, said David Slonina, the city’s deputy director of information technology. “It’s very self-contained,” he said.
Every other use of AI in city tech was incidental to what software did, Slonina said, such as in sorting email, running traffic lights or preventing viruses from getting into Somerville computer systems – which he resisted discussing specifically out of fear of hackers. “These AI technologies themselves are becoming targets for bad actors,” he said.
But in Somerville, Slonina said, “we don’t have any contracts or software products, to my knowledge, that are AI-specific, things like ChatGPT, Gemini, OpenAI.”
The report was accepted with reluctance by councillors who saw its limitations. “We’ve all seen the police department generate AI graphics for their communications,” councilor JT Scott said. “We’ve heard in multiple public meetings about uses of various AI tools to generate work materials.” He gave the example of engineers who used it to examine satellite photos of the city to help estimate permeable surface area.
It’s unavoidable and hard to identify and track, Slonina said – “They’re plugging AI in wherever they can plug it in.”
Humans make decisions
The driving force behind Thursday’s three AI-focused orders, councilor Jon Link, introduced the topic mindful of those complications. The law to be discussed doesn’t mean the city can’t use AI, but that it has to be used with an understanding of its biases and flaws – more for figuring out how to format an Excel document than to decide where a library branch might be placed, what artist might get a grant or “if a community member goes to prison,” in his examples.
The law also requires content made mostly by AI to be labeled as such, and that an annual report tell “us and the world” how much of the tech Somerville uses and how, Link said.
Transparency was a theme in the civil rights motion as well, with councilor Will Mbah emphasizing that AI’s aid in getting government to deliver services more efficiently “must be transparent, accountable and consistent with civil rights and privacy protections.”
“Residents should be able to know when it is being used materially to influence decisions affecting their housing assistance, city services, benefits, employment or other important opportunities,” Mbah said. “We should distinguish low-risk administrative uses from higher-risk application that could affect residents’ rights while ensuring appropriate human oversight.”
To be shaped in committee
The matters heading off to committee are acknowledged as imperfect, councilor Ben Wheeler said, but are in “good enough shape to open it to the close reads that we are eager to hear from members of the administration and members of the public, the rest of the council and advocacy groups.”
The lines to be drawn are tricky to define well, “and for me the guiding goal is clarity. We want the public and city staff to understand these limited requirements and the high-risk cases they apply to,” Wheeler said.
The resolution that passed, led by Link with Scott and councilor Naima Sait, was the simplest – it called just for “commonsense policies” for use of AI in creating media and in how City Hall and its departments are staffed.
“Since we are a city full of artists and wonderful creative people, if we need creative work for our publicly shared materials, whether that’s a flyer or a video, we should use those artists and not AI,” Link said. “The city should not use AI to replace workers – so no layoffs or closed positions specifically so that they can be filled by some sort of AI agent. I’m all for a city using AI, but it should be a force multiplier and not something that removes a person’s livelihood.”
