The report
Security desk
CISA added CVE-2026-114625, a SQL injection flaw in Arista VeloCloud Orchestrator, to its Known Exploited Vulnerabilities Catalog based on evidence of active exploitation. CISA encourages all organizations to adopt risk-based approaches and prioritize addressing vulnerabilities in its KEV Catalog.
- CISA
Primary source · U.S. public domain
Source published
Read by the newsroom
Planetary Risk desk
Asteroid 2026 UA1 passed 0.58 lunar distances from Earth on October 23, at 9.5 kilometers per second. The close-approach data lists 8 asteroid flybys within 10 lunar distances of Earth in the 7 days from October 23.
- NASA Jet Propulsion Laboratory source material
Data feed · reused under its publisher's terms (numeric facts; SSD asks for a citation; NASA's media guidelines)
Source published
Read by the newsroom
AI & Automation desk
Researchers evaluated six language models on their ability to generate synthetic survey respondents and found that all of them left human respondents uncovered when measured by respondent coverage gap. Every language model tested had a higher respondent coverage gap than a Gaussian copula fitted to the human data across the prompting and inference settings evaluated.
Earth & Sky desk
USGS lists 16 earthquakes of magnitude 4.5 or greater worldwide in the 24 hours to October 23 at 3:30 p.m. Central. The largest earthquake in that period is magnitude 5.7, 61 km WNW of Leni, Italy, on October 23 at 5:19 a.m. Central, at a depth of 10 kilometers.
- U.S. Geological Survey
Data feed · U.S. public domain
Source published
Read by the newsroom
Earth & Sky desk · closing story
The Global Meteor Network's daily summary generated October 23 at 9:55 a.m. Central lists 3,348 meteor trajectories measured by its cameras. The most active shower in the summary is the Orionids, with 1,006 meteors.
- Global Meteor Network
Data feed · CC BY 4.0
Source published
Read by the newsroom
Science desk
Researchers analyzed brain imaging data from 30 participants reading 108 sentences each to compare how transformer language models and other semantic models represent sentence meaning relative to human brain activity. Transformer models matched brain representations better than models that completely ignore word order, but all transformers poorly matched human brain representations of sentences requiring sensitivity to structural relations such as semantic roles. Transformers performed significantly worse than alternative models that were specifically built to capture the structural relations between words.
- eLife
Research · CC BY 4.0
Source published
Read by the newsroom
The standards box
Standards checked 14 claims and cut 1. Every claim is checked against its source; anything not supported is cut, never rewritten.
Corrections
None.