OpenAI, Anthropic and DeepSeek take AI safety to the UN Security Council

OpenAI, Anthropic and DeepSeek brought competing AI powers into the same United Nations Security Council briefing on September 23, warning that increasingly capable systems could improve themselves, evade human control and create international-security risks. OpenAI CEO Sam Altman also urged governments to establish benchmarks for measuring model capabilities and corporate safeguards.
Source

👉@sysadminoff

https://4sysops.com/archives/openai-anthropic-and-deepseek-take-ai-safety-to-the-un-security-council/

Microsoft’s Ink Canvas turns Surface handwriting into Copilot-ready content

Microsoft’s new Ink Canvas gives the refreshed Surface Pro more than a standard note-taking app: it combines handwriting, sketches, and annotations so Copilot can interpret users’ visual ideas. The feature adds a new layer to the refreshed Surface devices Microsoft announced recently, arriving preinstalled on the 12-inch Surface Pro 2nd Edition and in the Microsoft Store on October 13, 2026.
Source

👉@sysadminoff

https://4sysops.com/archives/microsofts-ink-canvas-turns-surface-handwriting-into-copilot-ready-content/

Qualcomm реализует поддержку чипов Snapdragon X2 в Linux

Компания Qualcomm объявила о передаче в основной состав ядра Linux изменений, обеспечивающих поддержку устройств на чипах Snapdragon X2, оснащённых CPU Qualcomm Oryon (6-18 ядер в зависимости от модели), AI-ускорителем Hexagon (80 TOPS) и GPU Adreno X2-90. Среди прочего будет подготовлен FastRPC-драйвер для NPU Hexagon и добавлена поддержка новых GPU Adreno в Mesa-драйверы Freedreno, Turnip и Rusticl.

👉@sysadminoff

https://www.opennet.ru/opennews/art.shtml?num=66334

Huawei’s UnifiedBus targets the networking bottleneck in agentic AI

Huawei is shifting its AI infrastructure pitch from faster accelerators to faster connections between them. At Huawei Connect 2026, the company introduced UnifiedBus alongside the Atlas 960E SuperPoD, claiming the architecture can reduce latency, increase bandwidth tenfold, and scale systems for massive agentic-AI workloads.
Source

👉@sysadminoff

https://4sysops.com/archives/huaweis-unifiedbus-targets-the-networking-bottleneck-in-agentic-ai/

📰 KDE developers' attempts at creating LLM guidelines are not going well

It seems the KDE developers want to allow some use of generative AI / LLMs, as they're working on a policy of guidelines for it.Read the full article on GamingOnLinux.

🔗 Source: https://www.gamingonlinux.com/2026/09/kde-developers-attempts-at-creating-llm-guidelines-are-not-going-well/

#kde

👉@sysadminoff

https://gamingonlinux.com/2026/09/kde-developers-attempts-at-creating-llm-guidelines-are-not-going-well

📰 Linux Adding New Interface For Reporting AMD SEV Status

A convenient addition that surprisingly hasn't been added to the Linux kernel until now is /sys/devices/system/cpu/sev/sev_status as a new sysfs interface for reporting the AMD Secure Encrypted Virtualization (SEV) status...

🔗 Source: https://www.phoronix.com/news/Linux-sysfs-AMD-SEV-Status

#amd #kernel #linux

👉@sysadminoff

https://phoronix.com/news/Linux-sysfs-AMD-SEV-Status

Cloud Range gives SOCs a proving ground for AI agents

Cloud Range, a cyber range platform for simulating enterprise security operations and adversary attacks, has launched an AI Validation Range and AI Readiness Framework designed to show whether security agents are safe to deploy—not merely whether they can complete a task. A security operations center (SOC) is the team and technology responsible for monitoring, detecting, investigating, and responding to cyber threats. Organizations can test agents against realistic attacks, compare their results with human defenders, and identify where human approval remains necessary.
Source

👉@sysadminoff

https://4sysops.com/archives/cloud-range-gives-socs-a-proving-ground-for-ai-agents/

Similarweb and Chartbeat reveal how AI is starving media sites of search traffic

New data from Similarweb and Chartbeat, reported by Axios, shows how AI-driven search and chat interfaces are cutting deeply into the referral traffic and engagement that once supported news websites. Google appears to be the main culprit: its AI Overviews answer more queries directly, reducing the need to visit original sources. Similarweb recorded just 47.6 million U.S. visits to about 100 major news sites in July, down from more than 70 million at the 2024 peak, while AI referrals remain too small to compensate.
Source

👉@sysadminoff

https://4sysops.com/archives/similarweb-and-chartbeat-reveal-how-ai-is-starving-media-sites-of-search-traffic/

Microsoft Teams can now block external bots before they enter meetings

Microsoft Teams now lets administrators automatically reject identified external bots instead of sending them to the lobby for manual approval. The control extends Teams’ meeting bot detection, helping organizations keep unauthorized AI notetakers, transcription services, and recording tools out of sensitive calls.
Source

👉@sysadminoff

https://4sysops.com/archives/microsoft-teams-can-now-block-external-bots-before-they-enter-meetings/

📰 Flatpak 1.18.3 Fixes Subsandbox Startup and Updates Meson Wrap Subprojects

Flatpak 1.18.3 Linux application sandboxing and distribution framework is now available for download with various bug and crash fixes for the stable Flatpak 1.18 series.

🔗 Source: https://9to5linux.com/flatpak-1-18-3-fixes-subsandbox-startup-and-updates-meson-wrap-subprojects

#flatpak #linux

👉@sysadminoff

📰 Classic physics-based browser game Happy Wheels gets revived on Steam

Happy Wheels is a game all about getting from A to B, and failing constantly while creating a rather amusing bloody mess.Read the full article on GamingOnLinux.

🔗 Source: https://www.gamingonlinux.com/2026/09/classic-physics-based-browser-game-happy-wheels-gets-revived-on-steam/

#steam

👉@sysadminoff

https://gamingonlinux.com/2026/09/classic-physics-based-browser-game-happy-wheels-gets-revived-on-steam

Claude Opus 5.5 cuts long coding-session costs with cheaper cache reads

Claude Opus 5.5 is designed for the way developers now use coding agents: longer sessions, larger contexts, and fewer interruptions. Anthropic says the model can cost about 40% less than Opus 5 on typical token-billed workloads, with the biggest savings coming from cached context and fewer wasted turns.
Source

👉@sysadminoff

https://4sysops.com/archives/claude-opus-5-5-cuts-long-coding-session-costs-with-cheaper-cache-reads/

Ubuntu переходит на ускоренный цикл устранения уязвимостей в пакетах с ядром Linux

Компания Canonical объявила о переходе на унифицированный 2-недельный цикл подготовки обновлений пакетов с ядром Linux с публикацией нового обновления каждую неделю. Цикл разработки обновлений сокращён в связи с увеличением интенсивности выявления уязвимостей в ядре Linux. Новый цикл формирования очередного корректирующего обновления пакета с ядром теперь будет запускаться каждую неделю без разделения на обычные обновления и обновления с устранением уязвимостей.

👉@sysadminoff

https://www.opennet.ru/opennews/art.shtml?num=66335

📰 New Linux Kernel Flaw Gives ARM64 KVM Guests Read-Write Access to Host Memory

A new flaw in the Linux kernel's KVM virtualization code for ARM64 processors can leave a freed piece of host memory exposed to a guest virtual machine on hosts with nested virtualization enabled.The bug, tracked as CVE-2026-89775, allows a guest to read and write host kernel memory, and the researcher who found it says it can be used to escape the guest and run code on the host machine.

🔗 Source: https://thehackernews.com/2026/09/new-linux-kernel-flaw-gives-arm64-kvm.html

#kernel #linux

👉@sysadminoff

📰 Alibaba Contributing $3M USD To Omarchy To Work On Making "Ideal" Agentic OS

The Omarchy project announced today that Alibaba Cloud will be contributing a total of three million USD to their foundation for helping development of this Linux distribution and working to make it an "ideal agentic operating system"...

🔗 Source: https://www.phoronix.com/news/Omarchy-Alibaba-3M

#linux

👉@sysadminoff

https://phoronix.com/news/Omarchy-Alibaba-3M

📰 Steam Tests New Pyrowave Codec on Linux and SteamOS

Valve is testing Pyrowave, a new low-latency Steam Remote Play codec available on Linux and SteamOS through the experimental SteamRT3 client.

🔗 Source: /

#linux #steam

👉@sysadminoff

https://linuxiac.com/steam-tests-new-pyrowave-codec-on-linux-and-steamos

📰 systemd 262 Released With AI/LLM Canary For Unreviewed Code, LUO Integration

Systemd 262 made it out today as the newest feature update to this Linux init system and service manager. As with most systemd releases, it's heavy on new features and functionality...

🔗 Source: https://www.phoronix.com/news/systemd-262

#linux #systemd

👉@sysadminoff

https://phoronix.com/news/systemd-262

📰 Raspberry Pi Disallows RAM Upgrades to Fight Fraud, Does it Make Sense?

A resurfacing firmware lock is catching out new buyers and tinkerers who want to revive dead hardware.

🔗 Source:

#raspberry

👉@sysadminoff

https://feed.itsfoss.com/link/24361/17469532/raspberry-pi-blocks-ram-upgrades

📰 Giada 1.6 Loop Machine Adds Time-Stretching Playback Mode for Sample Channels

Giada 1.6 open-source minimalistic and hardcore music production software has been released today as a major update that introduces several new features and improvements.

🔗 Source: https://9to5linux.com/giada-1-6-loop-machine-adds-time-stretching-playback-mode-for-sample-channels

#opensource

👉@sysadminoff

📰 SPEC Releases SPECviewperf 15.0.1 For Linux

SPEC back in May 2025 introduced SPECViewPerf 15 as the newest major release of this industry-standard workstation graphics benchmark. Sadly, the Linux support once again fell behind with SPECViewPerf 15 only debuting on Microsoft Windows. Today though Linux has caught up with SPEC releasing a Linux-native build of SPECViewPerf 15.0.1...

🔗 Source: https://www.phoronix.com/news/SPECviewperf-15.0.1-Linux

#linux

👉@sysadminoff

https://phoronix.com/news/SPECviewperf-15.0.1-Linux

📰 KDE's proposed LLM policy discussion gets locked after a fiery community clash

With the Linux kernel accepting LLM submissions, distro developers are beginning to ask themselves if AI code has a place within their own project. We've seen distros take both sides, with Gentoo and Void choosing the anti-AI option while Debian allowing only high-quality LLM-generated content. Now, it's KDE's turn to work out what level of LLM-made content is allowed, and the community has some strong thoughts from both camps.

🔗 Source: https://www.xda-developers.com/kdes-proposed-llm-policy-discussion-gets-locked-after-a-fiery-community-clash/

#debian #distro #kde #kernel #linux

👉@sysadminoff

https://xda-developers.com/kdes-proposed-llm-policy-discussion-gets-locked-after-a-fiery-community-clash