Monday, September 21, 2026 · Daily edition
SPP’s 90-day large-load path, the Fed’s buildout reading, and WHO’s AI-health ethics gap
Today’s ledger follows Southwest Power Pool’s conditional fast path for data centers and other giant loads, a Board of Governors staff note that still reads U.S. public data as an AI buildout rather than broad job displacement, NIST’s new concept note for trustworthy AI inside critical infrastructure, a WHO report on ethics oversight for AI-related health research, and Mongolia’s government-owned teacher assistant built on a national learning platform.
SPP’s large-load path: conditional service, parallel generation study, and a 90-day interconnection claim
What happened. Southwest Power Pool’s High Impact Large Load Integration page frames data centers, advanced manufacturing, and industrial facilities as the customers that need faster grid connection without hiding cost allocation or reliability tradeoffs. Two instruments sit at the center. Conditional High Impact Large Load service offers quicker study and interconnection, with the explicit possibility of curtailment during system stress. The High Impact Large Load Generation Assessment studies those loads and their supporting, often on-site generation in parallel. SPP says the process can deliver a path to interconnection agreements within 90 days. The page also carries a public Large Load Processes Q&A series, with materials and a transcript from 17 September 2026 and a next session on 15 October 2026. The HTML does not publish an independent megawatt or terawatt-hour census.
What to watch. After West Coast tariff clocks and Eastern ride-through fights, the Plains story is speed sold through conditional interconnection and parallel generation study. That is a reliability bargain, not another long-run electricity path.
Read the SPP High Impact Large Load page →
Fed staff note: public data still look like an AI buildout — younger workers feel it in slower hiring
What happened. Board of Governors staff economists Soto, Thieu, and Allen map public indicators from AI capabilities and costs through firm investment and adoption into productivity and labor in the FEDS Note The AI Buildout and the Economy (17 July 2026). On the page, aggregate unemployment remains moderate by historical standards, with possible compositional shifts. Youth unemployment and younger-cohort labor-force participation are named as series to watch if AI substitutes for entry-level tasks. Early evidence cited there: AI is affecting younger workers through slower hiring rather than outright layoffs. The authors’ bottom line is that the evidence still looks like a buildout phase, not the onset of broad-based displacement; labor-market impacts remain concentrated. Capability notes include METR agentic task-completion horizons in software engineering and machine learning roughly doubling every several months as of mid-2026, which could put a full workweek of agentic tasks within years as a technical possibility, not a cost-effective substitution claim. Investment notes treat hyperscaler capital spending as a demand proxy that still mixes non-AI spending; an Amazon year-end 2025 footnote says only about one-third of gross property, plant, and equipment was servers and networking. Census Business Trends and Outlook Survey uptake rises with firm size, while intensity can stay shallow even where adoption is broad. High-AI-exposure sectors show higher labor-productivity growth, but the page is careful that micro gains are not clearly adding up in the aggregate.
What to watch. After own-job versus whole-market perception splits, the jobs beat is a Board staff dashboard that still labels the United States a buildout story: investment is visible; mass pink slips are not. It is staff analysis, not FOMC policy and not a layoff census.
Read the Board FEDS Note →
NIST starts an AI RMF profile for trustworthy AI in critical infrastructure
What happened. NIST’s concept note for an AI Risk Management Framework profile on Trustworthy AI in Critical Infrastructure — created 6 April 2026 and updated 17 July 2026 — says critical infrastructure will rely more on AI across information technology, operational technology, and industrial control systems, and that those high-stakes settings need systems “worthy of trust.” NIST’s Information Technology Laboratory is launching the profile so operators can point to concrete risk-management practices and communicate trustworthiness requirements across AI and infrastructure lifecycles and supply chains. A Community of Interest is open for discussion drafts and feedback. The aim is more operator confidence to deploy AI agents and tools, plus clearer targets for vendors. Status on the page is concept note plus community, not a finished published profile and not a binding control overlay.
What to watch. The policy beat is how U.S. standards bodies shape trust requirements where AI touches critical infrastructure operators. That is a different instrument from content-marking codes and vulnerability-database modernization.
Read the NIST concept note →
WHO: ethics committees may lack the tools to oversee AI-related health research
What happened. On 21 September 2026, WHO announced the report Artificial Intelligence-related health research: ethics review and oversight, with recommendations for researchers, ethics committees, regulators, funders, and policy-makers. It was jointly developed by WHO research-ethics, science, and digital-health experts and is framed as a starting point for future standards, not a statute and not a bedside trial. The report sorts work into three categories: health research with data that uses AI; research with AI tools and technologies; and health research on AI tools and technologies. Oversight is meant to run from study design and ethics review through publication, regulation, and implementation. The page is blunt that existing ethics oversight may not be equipped for transparency, bias, fairness, accountability, privacy, and harms from rapid deployment. Research ethics committees remain central, but may need more expertise, training, and resources; funders, journals, data-governance bodies, professional societies, and regulators sit beyond one-project review. Much AI research and technology development remains concentrated in higher-income settings, so local leadership and capacity in lower- and middle-income countries are framed as necessary to avoid new exclusion.
What to watch. After discovery headlines and clinical risk models, the science beat is governance of how AI-health studies get reviewed. Capacity and equity are the stakes — not a new clinical endpoint.
Read the WHO announcement →
Mongolia’s Teacher Virtual Assistant: government-owned curriculum tools on a national platform
What happened. A World Bank Education blog on Mongolia describes the Teacher Virtual Assistant delivered through Medle, the country’s national learning platform. The first application helps teachers generate lesson plans and differentiated practice questions for Grades 1–5 mathematics from a government-owned curriculum dataset of more than 700 lesson plans and 4,900 practice questions. The architecture is three layers: digital public infrastructure such as unique student and teacher identifiers, interoperable information systems, connectivity, and governance; modular AI platforms that hold curriculum repositories and learner profiles, with the assistant as the first application; and financing or public-private partnerships so pilots are not stranded. Layer 1 is named as strengthened under a Global Partnership for Education System Capacity Grant plus Mongolia’s first Education Quality Standards Framework. Phased rollout is expected to produce evidence on teacher adoption and classroom practice. It is not a learning-outcomes randomized trial and not an enrollment census.
What to watch. After child-uptake snapshots and adult skills campaigns, the education beat is how a national platform puts curriculum-owned teacher tools on digital public infrastructure first. That is system design, not chatbot hype.
Read the World Bank Mongolia note →
Also in today’s ledger
• A 90-day conditional interconnection path is not a terawatt-hour census.
• A staff buildout dashboard is not a pink-slip count — and a concept note is not a finished control overlay.
• Ethics-oversight guidance is not a bedside trial — and a national teacher tool is not a learning-outcomes RCT.
|