Curated from knowledge graph (874 nodes, 926 edges) · All articles published within the last 7 days

• Microsoft Report Highlights Trends in Higher Education AI Adoption — About 9 in 10 students, educators, and education leaders say they have already used AI for school, according to Microsoft's third annual AI in Education Special Report, which consolidates survey results from more than 3,000 instructors, education leaders, and students in K–12 and higher education worldwide. More than half of education leaders now use AI daily, versus about a third of teachers and a quarter of students. The sharpest findings are perception gaps: 4 in 5 education leaders say their institution's AI guidance is clear, but half of students and teachers say that guidance is nonexistent or neutral — and while 7 in 10 leaders believe at least half of their students and instructors have had AI training, 77% of students and 53% of educators say they have received none. 🔗 Graph: Higher Ed AI, AI Adoption, AI Governance 📅 Published: 2026-09-11 📰 https://edtechmagazine.com/higher/article/2026/09/microsoft-report-highlights-trends-higher-education-ai-adoption 📌 Key takeaways: • The adoption question is settled; the support question is not — campus leaders should assume AI use is already near-universal and redirect strategy toward training cadence, tool governance, and classroom-level guidance, which is where the survey shows the widest gaps between leader perception and user experience. • The guidance-perception gap is a communication and co-design problem, not a drafting problem: if leadership believes policy is clear while students and faculty experience it as absent, the fix is auditing how AI guidance actually reaches people — before writing more of it. • The training gap is a service-demand forecast, with most students and educators asking for recurring, role-based training — so institutions should budget for ongoing AI literacy programs rather than one-off orientation sessions.

• Year-Round Education Requires Year-Round IT Support — As institutions expand hybrid, online, and year-round enrollment in response to the demographic cliff, EdTech Magazine examines how IT support models built around a traditional 9-to-5, fall-to-spring schedule leave nontraditional learners without help exactly when they are working. The piece draws on institutional practice: Southern New Hampshire University has invested heavily in knowledge management and self-service support, while Western Governors University is piloting AI-driven virtual agents in its contact center so students can get assistance outside staffed service-desk hours, with built-in escalation to a live person for anything the agents cannot resolve. The throughline is contributor Brad Hachez's recommendation to treat IT support as a learner-success function rather than a technical operation. 🔗 Graph: Higher Ed AI, AI Adoption 📅 Published: 2026-09-10 📰 https://edtechmagazine.com/higher/article/2026/09/year-round-education-requires-year-round-it-support 📌 Key takeaways: • Support hours are a retention issue: institutions pursuing adult, online, and year-round learners should evaluate service-desk coverage against when those students actually study — evenings and weekends — not against staff schedules. • AI virtual agents with human escalation are becoming the standard pattern for extending service-desk coverage without extending headcount; the agent-first model with built-in escalation to live staff is the reference architecture to pilot. • Knowledge management is the prerequisite investment: self-service and AI-assisted support are only as good as the knowledge base underneath them, so that work should be sequenced before deploying support agents.

• Rapidly scaling online storage to serve over 1 billion ChatGPT users — OpenAI's engineering team published a rare inside look at Habitat, the storage platform behind ChatGPT and its other products, which has grown from a simple Python library launched at DevDay 2023 into a distributed system handling more than 70 million requests per second, serving over 1 billion weekly users across nearly 40 regions, and storing more than 500 petabytes. The architecture breakdown covers the primitives any large multi-tenant service needs — caching layers, ACL policies, data residency, multi-tenancy isolation, and rate limiting — backed by Azure Cosmos DB, an internal store called Nanobase, and Valkey. The detail drawing the most attention: in Q2 2026, two engineers, working with Codex and GPT-5.5, rewrote the entire service in Rust, and that rewrite now handles 95% of production requests. 🔗 Graph: OpenAI, AI Adoption 📅 Published: 2026-09-11 📰 https://openai.com/index/scaling-storage-one-billion-users-part-one 📌 Key takeaways: • AI-assisted rewrites of critical infrastructure are now production-proven at frontier scale — a two-engineer Rust rewrite of a 70-million-requests-per-second service is the strongest public data point yet for what small platform teams can do with coding agents. • The platform's core primitives (per-tenant ACLs, data residency, isolation, rate limiting) double as a reference design for institutions building multi-tenant campus AI services: consumer scale forced OpenAI to solve the same problems a shared institutional platform faces. • The trajectory — Python prototype to load-bearing service to Rust rewrite — mirrors what campus-built AI services experience as usage grows; teams should design early prototypes with the assumption they will become mission-critical.

• The Agent Incident Registry: Toward Preventing Repeated AI Agent Failures — A new arXiv paper introduces the Agent Incident Registry (AIR), a source-linked catalog of publicly disclosed AI-agent incidents in which each record carries supporting evidence, a stable identifier, and structured labels for causal role, disclosure class, mechanism, and outcome. Analyzing the collection, the authors find that realized outcomes concentrate in "in-the-wild" and safety-failure records, while responsible disclosures and research demonstrations are overwhelmingly just demonstrated — and, in a deployment-analogue audit, InjecAgent's benchmark cases all occupy attacker-triggered surfaces even though the registry contains a distinct population of no-adversary safety failures. The registry is explicitly scoped for source-grounded case retrieval and evaluation-scope auditing rather than failure-rate or control-efficacy estimation. 🔗 Graph: AI Security, Agentic AI 📅 Published: 2026-09-12 📰 https://arxiv.org/abs/2609.11030 📌 Key takeaways: • A substantial share of real agent failures are not attacks: governance for campus AI agents should extend beyond prompt-injection threat modeling into reliability engineering, staged rollouts, and internal incident review. • The taxonomy is directly reusable — institutions deploying agents can adopt the same labels (causal role, disclosure class, mechanism, outcome) for internal incident postmortems, making them comparable to the public registry and to vendor security evaluations. • Security teams benchmarking agent tooling should audit which failure classes their tests actually cover, since attacker-triggered suites can score well while missing the no-adversary safety failures that dominate realized harm.

• Top Scientists Lead Growing Exodus of Chinese Academics From U.S. — Inside Higher Ed relays Times Higher Education reporting that Chinese researchers are leaving U.S. universities in growing numbers, driven by a mix of push factors — a more hostile U.S. foreign policy, funding uncertainty, and heightened scrutiny of China — and pull factors, notably larger research budgets in China. The exodus is a structural trend that has been accelerating since the 2010s, but 2026 has seen more than a dozen very high-profile Chinese scientists leave prestigious U.S. roles for top positions in China, and returnee academics are often among the strongest research performers. The piece situates the departures within deepening U.S.–China technological and scientific rivalry. 🔗 Graph: Higher Ed AI, AI Governance 📅 Published: 2026-09-11 📰 https://www.insidehighered.com/news/faculty/research/2026/09/11/top-scientists-lead-exodus-chinese-academics-us 📌 Key takeaways: • Research universities should treat talent retention as an AI-era strategic risk: the fields losing the most prominent researchers are precisely those where U.S.–China competition is sharpest, and departures of top performers compound through recruiting networks and graduate pipelines. • The levers are partly institutional: funding stability, visa predictability, and how compliance and scrutiny requirements are administered day to day are areas where university leaders can reduce push factors even while national policy is beyond their control. • For research-computing and IT leaders, sustained faculty departures can shift infrastructure demand quickly — capacity plans and shared-research investments should be stress-tested against talent flows, not just enrollment projections.