#118: Impact Signals #118 — Opioid Settlement Cash Bought AI Plate Readers, Punjab Pilots a Ten-Disease CT Scanner, Lesotho Puts Every Farmer in One Registry
AI for Impact Daily Briefing, August 31, 2026
Top Stories
Nineteen Police Agencies in Ten States Bought AI Plate Readers With Opioid Settlement Money
Mother Jones traced opioid settlement spending to automated license plate readers and found 19 law enforcement agencies across 10 states that put more than 920,000 dollars of settlement money into AI camera systems from vendors including Flock and Axon. The settlements total more than 50 billion dollars paid by pharmaceutical firms over an 18-year distribution schedule, awarded to communities harmed by an addiction crisis that killed roughly 806,000 people between 1999 and 2023. Sunnyside, Washington approved 24,000 dollars for a one-year lease on a mobile Flock trailer in a town with one of the highest overdose death rates in the state and no medical detox centre; the Terrebonne Parish Sheriff's Office in Louisiana was reported at 465,000 dollars on Flock cameras. West Virginia jurisdictions named include Monongalia, Ohio, Randolph and Upshur counties and the cities of Weirton and Martinsburg. Robyn Oster, director of policy at the Partnership to End Addiction, put the objection in one line: "Whatever funds you're spending on a camera, you're not spending on getting someone into treatment." Why it matters: This is the first time the plate-reader buildout has been traced to a specific, legally earmarked public health fund with a dollar figure attached, which converts a general privacy argument into a budget argument that county commissions have to answer in public. The settlement money runs on an 18-year schedule and the contracts named here are leases and multi-year deals rather than one-time buys, so the same line items return for renewal.
Punjab Approves a Pilot of an AI CT Scanner That Screens for Ten Diseases in Two Minutes
Punjab Chief Minister Maryam Nawaz Sharif gave in-principle approval on August 31 to pilot an AI-powered multi-disease screening system across the province and directed provincial health authorities to prepare the pilot, following a briefing from Alibaba Cloud executives during her China visit. The system runs on non-contrast CT, so it needs no injected contrast agent, and the claim being piloted is that one scan can flag up to ten conditions in about two minutes: pancreatic, liver, gastric, oesophageal and colorectal cancers, acute aortic syndrome and cardiovascular risk, fatty liver disease and osteoporosis, plus fat and muscle assessment. Officials framed it as complementing conventional diagnosis rather than replacing it. Punjab is Pakistan's largest province by population, so a tool that runs on existing CT hardware is a throughput decision as much as a clinical one. Why it matters: A provincial government has committed to a pilot with a named modality and a named vendor relationship, which moves this out of the memorandum-of-understanding category that most government AI health announcements stay in. What is verifiably next is the pilot design the health department was told to prepare: how many sites, whose scanners, and whether the ten-condition claim is validated on Punjabi patients before it is scaled.
Lesotho Launches a National Farmers Portal Covering All 10 Districts
Lesotho's Ministry of Agriculture, Food Security and Nutrition launched the Lesotho Farmers Portal on August 30, the country's first national platform holding farmer and land information in a single registry. Minister Selibe Mochoboroane led the launch with Vodacom Lesotho as launch partner; before it, farmer records sat in paper files, spreadsheets and separate district systems that did not connect. Field officers in all 10 districts will use the portal to register farmers and land parcels and to record household details, farming activities, livestock, assets and support services. The build is deliberately unglamorous: Joget DX Enterprise as the low-code platform, the International Telecommunication Union supplying the digital-government methodology, GovStack supplying the Registration Building Block specification, and the World Food Programme supporting implementation including the software licence. The ministry keeps ownership of the registry, the data and the process. Why it matters: A national farmer registry is the layer anticipatory cash, subsidy delivery and crop insurance all sit on, and its absence is the usual reason those programmes target by district instead of by household. What is verifiably next is coverage: the portal exists in all 10 districts, but the number of farmers actually registered is the figure that determines whether the next drought response can reach people individually.
SenseNet Buys N5 Sensors' Wildfire Division, Bringing One Stack Over 300 Million Acres
SenseNet, a Vancouver company founded in 2019, acquired the wildfire division of N5 Sensors; deal terms were not disclosed. Its platform combines ground sensor networks, AI long-range cameras, satellite and drone feeds and predictive analytics, and the company says its sensors and cameras now monitor more than 300 million acres for fire starts. About 65 percent of revenue is private sector, heavy industry including construction, mining and telecommunications with Rogers named as a customer, and about 35 percent is public sector, including fire departments in California and the US Department of Homeland Security. SenseNet runs roughly 10 million Canadian dollars in annual revenue with just over 50 staff and expects 100 by year-end. Chief executive Hamed Noori said the acquisition is meant to build a full-service wildfire mitigation business aimed at underserved US markets, with a US office planned. Why it matters: Wildfire detection has been consolidating from single-modality vendors (cameras, or sensors, or satellites) into one stack, and this is a detection company buying the ground-sensor half rather than raising to build it. For fire agencies it means fewer procurement counterparties and a single vendor holding both the smoke-detection cameras and the pre-smoke chemical sensors, which is a resilience question as much as a coverage one.
San Jose's New Shelter Dashboard Publishes the Outcome It Cannot Measure
San Jose launched an internal Shelter System Dashboard in August, published August 28, with a public version expected in the autumn. It updates monthly and tracks shelter use, length of stay, demographics, people housed, programme exits and destinations, exits without notice, and participant deaths. The city counts 6,503 residents experiencing homelessness; California's statewide count fell from more than 187,000 in 2024 to 181,934 in 2025, a 2.8 percent decrease. Erik L. Soliván, director of the city's Housing Department, said San Jose cannot reliably report whether people are still housed 18 months after exiting a programme because that data is split between city systems and Santa Clara County systems: "This measure cannot currently be reported reliably." Why it matters: Cities are increasingly asked to justify shelter spending on outcomes rather than bed-nights, and San Jose has now published, inside its own dashboard, the specific data-sharing boundary that makes the outcome number unavailable. What is verifiably next is the public release in the autumn and whether the 18-month housing-retention measure appears in it.
Upcoming Events & Opportunities
CyberSafe Foundation with SANS Institute, AI Security Fellowship 2026 (Funding)
- Amount: fully funded nine-month training and certification pathway (GIAC AI Platform Security and GIAC AI Security Automation Engineer)
- Deadline: September 10, 2026, 12:00 PM WAT
- Eligibility: African women over 21, English-speaking, resident in an African country, at least three years of hands-on cybersecurity experience; no prior AI experience required
- Programme dates: selection feedback September 30, 2026; orientation October 21, 2026; training begins October 26, 2026; completion July 30, 2027
- Apply: cybersafefoundation.org
Active Disaster Monitoring (GDACS/OCHA)
- Nepal and Tibet:** Bhote Koshi glacial flood, onset August 26, ongoing. GDACS Orange flood alert with 527 deaths in the GDACS record as of its August 28 update; news reporting through August 31 puts the toll higher, in the 700 to 900 range with thousands still listed missing, figures attributed to Nepali authorities and not confirmed on a primary page. Search and rescue continues, including workers trapped in hydropower tunnels; fresh flood alerts were issued as Bhote Koshi flow rose again on August 30. Source: gdacs.org.
- India:** Flooding, onset August 9, ongoing. GDACS Orange flood alert; 20 deaths and 240,570 displaced in the GDACS record as of its August 30 update. Source: gdacs.org.
- China:** M5.0 earthquake, August 28, 05:13 UTC, depth 10 km. GDACS Orange alert; 15.8 million people within 100 km. Ongoing GDACS monitoring. Source: gdacs.org.
- Somalia:** M6.0 earthquake, August 27, 145 km north of Caluula. USGS alert level green, offshore, no damage reports. Monitoring only. Source: earthquake.usgs.gov.
- Timor-Leste:** M6.0 earthquake, August 24, 215 km NNE of Lospalos. USGS alert level green. Monitoring only. Source: earthquake.usgs.gov.
- Note: only major or ongoing-major disasters are featured; low-severity alerts are excluded per the major-only bar.
Sources: See individual stories above for full attribution.