Cambridge, Massachusetts, the city Harvard and MIT call home, is asking residents to help decide the future of ShotSpotter. The method is unusual — a group of ordinary residents will be selected by lottery for a jury-like review of the technology. Cambridge shut the system off in May over concerns about its reliability, privacy and who could access its data. But a fatal shooting less than two months later prompted the city to reverse course.
Xavier Bautista, a 32-year-old city worker, was shot around 4:30 a.m. July 4. His body was not discovered until about an hour later, when a passerby called 911. The killing prompted calls to restore ShotSpotter. Less than a month later, on Aug. 3, the City Council met at City Hall and voted 6-3 to reactivate it and create the resident assembly.
The technology will remain on while the city manager has 90 days to convene the assembly and return with its recommendations. The City Council will then decide whether to continue using ShotSpotter. SoundThinking, the company behind ShotSpotter, welcomed the vote in a statement to The Boston Globe and said it remained committed to working with Cambridge on public-safety measures that respect privacy and civil liberties.
AI Governance
One of the clearest public plans for prosecutorial AI comes from an unlikely place: the Republic of Maldives, a small island country in the Indian Ocean with roughly as many citizens as Cleveland has residents. On Aug. 11, the country's national prosecution authority put up a website laying out the plan in unusual detail.
Called PGO 2.0, the plan makes bright-line rules for where AI belongs in prosecutorial work. AI may analyze law, prepare draft charges and organize evidence; it may not decide whether a person should be charged. Further, the rules state case data cannot be entered into public AI systems, and no tool can go live until it is covered by policy and placed on an official register.
The Maldives is unlikely to set the global standard for prosecutorial AI. More sophisticated governance standards already exist in Portugal, Canada, England and Wales and Buenos Aires. Together, the five plans were published across four continents in just 14 months. The takeaway? Prosecution services around the world are now writing AI into their operating rules.
Flock’s newest product may be changing what it means to investigate a case. (WIRED)
Five federal agencies warn attackers are using AI to target industrial controllers. (The Record)
The FBI could spend up to $88 million building out its own AI compute. (FedScoop)
Montana is part of the first wave of courts trying to write specific rules for evidence accused of being an AI fake. (Montana Supreme Court)
Judges remain immune from suit even when accused of letting AI do the judging, a federal court in Nevada said. (U.S. District Court)
