Friday, September 18, 2026 · Daily edition
Boston Fed worker fear, PJM ride-through after a 4,000 MW drop, and FDA keeps radiology CAD on 510(k)
Today’s ledger follows a Boston Fed household survey where workers’ own-job AI-loss fear roughly doubled, a PJM proposal for large-load ride-through after nearly 4,000 MW of data-center load disconnected in northern Virginia, an FDA order denying a 510(k) exemption for radiology CAD and triage software, a Nature Medicine comment that clinical-AI trust needs prospective real-world evidence, and a UNESCO Strategy Lab on scaling offline AI and simulation learning where connectivity is thin.
Boston Fed: workers’ own-job AI-loss fear roughly doubled
What happened. The Federal Reserve Bank of Boston published Current Policy Perspectives 26-8 on workers’ views of AI productivity and job-loss risk, drawing on two New York Fed Survey of Consumer Expectations modules of about 1,300 U.S. household heads in December 2024 and December 2025. The share concerned about losing their own job to AI nearly doubled, from 5 percent at end-2024 to just over 10 percent at end-2025. Broader negative outcomes — job loss, lower wages, or looking for alternatives while unable to adapt — rose from about 11 percent to 18 percent. In 2025, about 10 percent worried about their own job, while 60 percent still expected AI-related layoffs or fewer workers in their industry. Own-job concern was highest in consumer services (23 percent), leisure (21 percent), and firm services (15 percent). Among the 6 percent of AI-exposed workers who reported the strongest productivity gains, 14 percent said they were more likely to ask for a raise because of AI.
What to watch. This is what workers themselves report, not a payroll census and not a production-function model. Fear is rising, industry cuts still feel more likely than personal ones, and the small group that feels big productivity gains is also the group more willing to ask for a raise.
Read Boston Fed CPP 26-8 →
PJM proposes ride-through after nearly 4,000 MW of data-center load dropped off
What happened. PJM Inside Lines reported a Planning Committee proposal, announced 8 September 2026, for new reliability rules so large computational loads — data centers and crypto-mining included — stay connected during normally cleared transmission disturbances. The trigger was a 22 July northern-Virginia event in the Dominion zone, when nearly 4,000 MW of data-center load unexpectedly disconnected from PJM and switched to on-site backup generation. Operators managed load-generation imbalance plus spikes in transmission voltage and system frequency. The grid avoided unresolvable reliability impacts that day, but PJM System Operations called it the third separate measurable event in two years and said the pattern is unsustainable. The proposal would set minimum voltage and frequency ride-through for existing and future large loads, with a transition plan for facilities already online. It also answers an 18 June FERC order to identify gaps in evaluating large-load reliability impacts, with proposed solutions due to FERC in November.
What to watch. After a stretch of queues, forecasts, and voluntary flexibility coalitions, the environment story is whether large AI loads must stay on the system through normal disturbances. This is proposed ride-through design, not another 2030 electricity-share path.
Read the PJM Inside Lines report →
FDA denies a 510(k) exemption for radiology CAD and triage software
What happened. The Federal Register published final order 2026-19074, effective 17 September 2026, denying a petition that sought a partial exemption from 510(k) premarket notification for radiology computer-aided detection and diagnosis devices and computer-aided triage and notification software. The petition had covered radiological computer-assisted diagnostic software for lesions suspicious of cancer, medical image analyzers, triage and notification software, and related CAD tools when petition conditions were met. FDA published notice of the petition in December 2025, told the petitioner of the denial in April 2026, and has now published the determination as a final order under the FD&C Act.
What to watch. This is a concrete device-gatekeeping decision: radiology CAD and triage software stay on the 510(k) path rather than getting an off-ramp. It is not a new clearance and not the still-open generative-AI medical-device discussion paper, whose comments remain due 19 October.
Read Federal Register order 2026-19074 →
Nature Medicine: clinical-AI trust cannot be benchmarked into existence
What happened. Nature Medicine published a comment arguing that trust in clinical artificial intelligence cannot be benchmarked into existence. It has to be earned through rigorous prospective studies in real-world clinical settings, where the hardest lessons often concern the humans and systems around the AI, not the model itself. The author block includes Google Research, Google DeepMind, Beth Israel Deaconess / Harvard, Included Health, and Stanford. Competing interests note Alphabet funding and possible Alphabet / Included Health equity.
What to watch. After a run of methods and simulation papers, the science beat is the bar for trust: prospective real-world evidence, not another leaderboard score. This is an evidence-standard argument, not a patient-outcome trial.
Read the Nature Medicine comment →
UNESCO: scale AI and simulation learning beyond the pilot where connectivity is thin
What happened. UNESCO published a Strategy Lab note from Digital Learning Week 2026 in Paris, organized by the UNESCO Regional Office for Southern Africa with Nudle. The focus is low-resource higher education: limited connectivity, uneven infrastructure, constrained budgets, educator readiness, and unequal access. The named vehicle is TECH SPARK Africa, with examples of simulation-supported learning, offline AI, and cross-device approaches where specialized equipment or reliable internet is limited. The core claim is that scaling digital transformation requires more than scaling technology — infrastructure and connectivity plus educator capacity, curriculum alignment, accessibility, institutional ownership, and long-term sustainability. The youth-centered line is that value is whether learning and opportunity improve, not how sophisticated the stack looks.
What to watch. The education beat is whether AI and simulation can work offline and at scale in low-resource higher education. This is implementation conditions, not another connected-campus pilot announcement.
Read the UNESCO Strategy Lab note →
Also in today’s ledger
• Worker fear is not a payroll census.
• Ride-through design is not a TWh path — and a 510(k) denial is not the GenAI discussion paper.
• An evidence-standard comment is not a bedside trial — and offline scale conditions are not a learning RCT.
|