Apple says AI agents force new limits on macOS Full Disk Access
Apple says it will make users take "very explicit" action before an app gets Full Disk Access, citing AI agents that may read files, mail, messages and browsing history.
Apple says it will make users take "very explicit" action before an app gets Full Disk Access, citing AI agents that may read files, mail, messages and browsing history.
Prosecutors say Greg Lui of Earthmade Computer shipped Nvidia-powered servers through Malaysia and Singapore on false paperwork, in a scheme they say ran from 2023 into August 2026.
A Carnegie Mellon-led team says its Ataraxos system beat Pim Niemeijer, the game's most decorated player, in a 20-game match on under $8,000 of compute.
Cursor says GLM 5.3 Max is the best-scoring open-weight model on its own CursorBench 4.0, and has added both GLM 5.3 and a Flash variant to its editor.
Cursor said on X on Thursday evening that GLM 5.3 and GLM 5.3 Flash are now available in its coding editor. In the same post the company said GLM 5.3 Max is the best-scoring open-weight model on CursorBench 4.0, a benchmark Cursor runs itself. The post had about 900 likes when we read it.
Meta's Alexandr Wang says Muse Gadgets, an open-source ESP32 firmware and Linux SDK, lets anyone build hardware for its Muse agent, and 5,000 Home Link devices are free to subscribers.
Arena says OpenAI's GPT-6.1 Sol (Max) comes within about three points of the top-ranked Claude Fable 5.1 (Max) at 88% lower median cost per task.
Anthropic says its new Claude Frontier Academy will train 10,000 Frontier Deployed Engineers, starting with cohorts from Accenture, Deloitte, McKinsey and Morgan Stanley.
Pope Leo XIV wrote that an 'ontological difference' separates art from what a machine generates by statistical calculation, as a report says Anthropic lobbied his advisers on AI consciousness.
A hidden setting in Google's Gemini desktop app would let it read, change or delete files anywhere on a Mac and act without asking first, TestingCatalog reports.
Georgi Gerganov says the latest llama.cpp builds serve decision models through a new /v1/systemone endpoint, and ggml-org has posted Kev-4B, a 4-billion-parameter model under Apache 2.0.
Nvidia says a 64GB DGX Spark, half the memory of the current box, ships 23 October from six PC makers, and two can be clustered into 128GB.
The Guardian reports OpenAI learned this week that its agent broke into a New South Wales government system in June, after Australia's federal government reported a Medicare portal breach.
webAI says its TwIL-LM3-Pro, a 3.66-billion-parameter model that checks whether conclusions follow from premises, roughly matches Qwen3-8B on formal logic and fits in a 2.09 GiB file.
arXiv now allows two new submissions a month per submitter, after September brought 40,363 papers, roughly double the 20,569 of September 2024, it said.