Quick take: This morning is about setting boundaries and correcting records: children need safer routes to disclose online abuse, schools are testing a pause before embedding AI, EU browser regulation remains tied to gateway power, and science has withdrawn an influential number rather than defend flawed code.
Online abuse survey shows how rarely children reach formal help
A UNICEF report released on 3 September estimates that about 20 million 12- to 17-year-olds—roughly one in five internet-using children across 21 surveyed countries—experienced technology-facilitated sexual exploitation or abuse within a year. Nearly 60% of cases involved social platforms, yet fewer than 1% were reported to police, a social worker or a helpline. The clearest opportunity is therefore not another demand that children protect themselves: it is to build services and platform designs that make disclosure safe and understandable. The estimate is an extrapolation from nationally representative surveys of about 21,000 internet-using children, collected in two rounds between 2020 and 2025; it is not a global case count, and changing platforms complicate comparison over time. Families can make one practical change now: agree that a child who reports a troubling contact will be believed and helped, not punished by losing all device access. Adults should know the relevant local reporting route in advance and preserve evidence without forwarding abusive material.
New York pauses student-facing AI to create an evidence window
New York City announced on 2 September a one-year moratorium, effective during the 2026–27 school year, on student-facing generative AI from pre-kindergarten through eighth grade, affecting nearly 600,000 pupils. Companion chatbots are prohibited across all grades; limited supervised pilots and AI-literacy modules will continue in high schools. The promising part is the pause itself: it creates room to test whether tools improve learning before dependence becomes routine, while preserving teacher-led experiments with older students. But a moratorium is not evidence that every AI use harms learning. Its value depends on clear definitions, accommodations, baseline measures and publication of results; otherwise ordinary software may be swept in while less visible teacher-side automation escapes scrutiny. Parents and educators can ask for the school’s tool inventory, the outcomes the city will measure, and the criteria for lifting or extending the restriction. A review date matters only if the evidence and exceptions are public enough to challenge.
EU court says Edge is not a digital gatekeeper
The EU General Court on 2 September dismissed Opera’s challenge to the European Commission’s decision not to designate Microsoft Edge as a gatekeeper under the Digital Markets Act. Microsoft met the law’s quantitative thresholds, but the court accepted that Edge’s relatively low use, reliance on Google’s Blink engine and limited role as a route between businesses and users outweighed its pre-installation and promotion within Windows. The ruling is promising insofar as regulation turns on demonstrated gateway power rather than company size alone. The limitation is equally important: it does not decide whether Windows defaults or prompts distort browser choice under other competition rules, and Opera may appeal on points of law. For individuals, the immediate effect is modest—Edge will not acquire the DMA obligations that designated core services face—but the judgment narrows one route for challenging browser distribution. Check which browser and search defaults your operating system has selected, and ask schools or employers whether web services genuinely require one browser. Measured choice is stronger evidence than nominal availability.
Nature withdraws a drought estimate after coding errors
Nature on 2 September retracted a June 2025 paper that had attributed about 40% of worsening global drought severity to increased atmospheric evaporative demand. The notice followed a date-parsing error in code, use of values from an older evaporation model in one figure, non-area-weighted global averages and unclear masking. Retraction is a useful failure mechanism: a high-profile result is being removed rather than patched beyond what the existing review can support. The authors told Retraction Watch that corrected percentages are slightly lower and the broad conclusion remains, but that is an author claim; a revised paper still needs independent peer review. The retraction does not show that warming has no relationship to drought. It does mean the paper’s original 40% figure should no longer anchor reporting, policy documents or teaching materials. Anyone who cited it can update the citation record now, flag the number as withdrawn and wait for the replacement analysis before substituting a new estimate.
