Berkeley report maps human rights risks of LLMs in three fields
A Human Rights Center study based on 56 interviews across 24 countries finds privacy and freedom-of-thought risks as law, journalism and education adopt large language models.
Author: TechUpscale Newsroom
Arm details role in NVIDIA’s Open Agent Safety Platform
Arm says its CPUs run AI agents while Arm-based DPUs such as NVIDIA BlueField-4 police them from a separate trust domain.
Cloudflare adds WebMCP support to its Kitesurf agent browser
Cloudflare says its Workers-based browser for AI agents now supports WebMCP tools, passes over 730,000 Web Platform Tests subtests and runs inside the terminal.
Microsoft Research Asia – Singapore marks first year of AI lab
Microsoft says its first Southeast Asian research lab has started nine new university projects and tied its AI agenda to Singapore's national strategy.
Cloudflare brings native Rust and Tokio to Workers via Emscripten
Cloudflare says an experimental preview lets Rust code built on Tokio, including real TCP sockets, run on Workers through a new wasm-bindgen Emscripten target.
NASA targets uncrewed Starliner-1 flight, crew return by 2028
NASA says Boeing's Starliner will fly uncrewed as soon as December to validate thruster fixes, with astronauts not expected back aboard until Starliner-2 in 2028.
GitHub Security Lab finds 24 Android bugs with AI taskflow agent
GitHub says custom LLM audit taskflows uncovered location tracking in OsmAnd and a Wikipedia app account takeover, but human review is still needed to judge severity.
NASA Swift boost mission ends without raising observatory’s orbit
NASA says Katalyst Space's LINK servicing spacecraft hit communications and attitude control problems and re-entered on Sept. 25 without grappling the Swift observatory.
AWS to cut European Sovereign Cloud off from global backbone Oct. 24
Amazon says the Brandenburg-based region will run for several hours without its global network backbone to produce evidence of EU operational independence.
Microsoft details NeedyMantis malware in targeted intrusions
Microsoft Threat Intelligence says the modular post-compromise framework has hit telecom firms, universities, nonprofits and government contractors since at least October 2025.