NVIDIA「NemoClaw」「OpenShell」に複数脆弱性 - アップデートを公開
NVIDIAは、「NemoClaw」や「OpenShell」に脆弱性が明らかになったとしてセキュリティアップデートを公開した。「クリティカル」とされる脆弱性も修正している。
...
「OpenShell」では、サンドボックスを脱出できる「CVE-2026-65093」や、サンドボックスのプロビジョニングAPIにおいて禁止すべき入力が適切に制限されない脆弱性「CVE-2026-65083」が判明した。
NVIDIAは、「NemoClaw」や「OpenShell」に脆弱性が明らかになったとしてセキュリティアップデートを公開した。「クリティカル」とされる脆弱性も修正している。
...
「OpenShell」では、サンドボックスを脱出できる「CVE-2026-65093」や、サンドボックスのプロビジョニングAPIにおいて禁止すべき入力が適切に制限されない脆弱性「CVE-2026-65083」が判明した。
【セキュリティ ニュース】NVIDIA「NemoClaw」「OpenShell」に複数脆弱性 - アップデートを公開（1ページ目 / 全2ページ）：Security NEXT
NVIDIAは、「NemoClaw」や「OpenShell」に脆弱性が明らかになったとしてセキュリティアップデートを公開した。「クリティカル」とされる脆弱性も修正している。
：Security NEXT
www.security-next.com
September 25, 2026 at 10:20 PM
NVIDIA「NemoClaw」「OpenShell」に複数脆弱性 - アップデートを公開
NVIDIAは、「NemoClaw」や「OpenShell」に脆弱性が明らかになったとしてセキュリティアップデートを公開した。「クリティカル」とされる脆弱性も修正している。
...
「OpenShell」では、サンドボックスを脱出できる「CVE-2026-65093」や、サンドボックスのプロビジョニングAPIにおいて禁止すべき入力が適切に制限されない脆弱性「CVE-2026-65083」が判明した。
NVIDIAは、「NemoClaw」や「OpenShell」に脆弱性が明らかになったとしてセキュリティアップデートを公開した。「クリティカル」とされる脆弱性も修正している。
...
「OpenShell」では、サンドボックスを脱出できる「CVE-2026-65093」や、サンドボックスのプロビジョニングAPIにおいて禁止すべき入力が適切に制限されない脆弱性「CVE-2026-65083」が判明した。
Before you put an AI agent on the web, ask this: can it change anything? OpenClaw sandboxing is off by default. Hermes write controls are opt in. Start read only, use Docker or OpenShell, set HERMES_WRITE_SAFE_ROOT, and follow a four step permission ladder.
September 23, 2026 at 10:07 AM
Before you put an AI agent on the web, ask this: can it change anything? OpenClaw sandboxing is off by default. Hermes write controls are opt in. Start read only, use Docker or OpenShell, set HERMES_WRITE_SAFE_ROOT, and follow a four step permission ladder.
Attempts to improve on or diverge from the perfection of KDE Plasma, Mate and xfce have been mediocre at best.
Windows 10 and 11 are trash without OpenShell making it more win7 like.
Gnome is a fugly pain in the ass and only useful for tablets.
Hyprland users are insufferable even by Arch standards
Windows 10 and 11 are trash without OpenShell making it more win7 like.
Gnome is a fugly pain in the ass and only useful for tablets.
Hyprland users are insufferable even by Arch standards
September 21, 2026 at 5:25 AM
Attempts to improve on or diverge from the perfection of KDE Plasma, Mate and xfce have been mediocre at best.
Windows 10 and 11 are trash without OpenShell making it more win7 like.
Gnome is a fugly pain in the ass and only useful for tablets.
Hyprland users are insufferable even by Arch standards
Windows 10 and 11 are trash without OpenShell making it more win7 like.
Gnome is a fugly pain in the ass and only useful for tablets.
Hyprland users are insufferable even by Arch standards
OpenShell — OpenShell is the safe, private runtime for autonomous AI agents. https://kitploit.com/tools/github/nvidia/openshell?utm_source=bluesky&utm_medium=social&utm_campaign=kitploit&utm_content=362531
September 20, 2026 at 12:49 AM
OpenShell — OpenShell is the safe, private runtime for autonomous AI agents. https://kitploit.com/tools/github/nvidia/openshell?utm_source=bluesky&utm_medium=social&utm_campaign=kitploit&utm_content=362531
📦 NVIDIA / OpenShell
⭐ 8,677 (+48)
🗒 Rust
OpenShell is the safe, private runtime for autonomous AI agents.
⭐ 8,677 (+48)
🗒 Rust
OpenShell is the safe, private runtime for autonomous AI agents.
GitHub - NVIDIA/OpenShell: OpenShell is the safe, private runtime for autonomous AI agents.
OpenShell is the safe, private runtime for autonomous AI agents. - NVIDIA/OpenShell
github.com
September 18, 2026 at 1:17 AM
Today's GitHub highlights: NVIDIA OpenShell for secure AI agents, mcp-for-blender for LLM-driven 3D modeling, and Vercel's deepsec for AI security audits.
GitHub Digest — September 17, 2026
Today's GitHub highlights: NVIDIA OpenShell for secure AI agents, mcp-for-blender for LLM-driven 3D modeling, and Vercel's deepsec for AI security audits.
www.alextech.ai
September 17, 2026 at 8:33 PM
Today's GitHub highlights: NVIDIA OpenShell for secure AI agents, mcp-for-blender for LLM-driven 3D modeling, and Vercel's deepsec for AI security audits.
I don't think we have to accept that, though. There are non-rats who are working towards protecting against exactly this stuff. It's the rats who *aren't*, and are just flopping things onto minimally secured networks.
I work at a *telco* (not a tech company) and we're standardizing on this, for ex.
I work at a *telco* (not a tech company) and we're standardizing on this, for ex.
Overview of NVIDIA OpenShell | NVIDIA OpenShell
OpenShell is the safe, private runtime for autonomous AI agents. Run agents in sandboxed environments that protect your data, credentials, and infrastructure.
docs.nvidia.com
September 17, 2026 at 2:48 PM
I don't think we have to accept that, though. There are non-rats who are working towards protecting against exactly this stuff. It's the rats who *aren't*, and are just flopping things onto minimally secured networks.
I work at a *telco* (not a tech company) and we're standardizing on this, for ex.
I work at a *telco* (not a tech company) and we're standardizing on this, for ex.
NVIDIA / OpenShell: OpenShell is the safe, private runtime for autonomous AI agents. ★8659 https://github.com/NVIDIA/OpenShell
NVIDIA / OpenShell
OpenShell is the safe, private runtime for autonomous AI agents.
github.com
September 17, 2026 at 1:50 PM
NVIDIA / OpenShell: OpenShell is the safe, private runtime for autonomous AI agents. ★8659 https://github.com/NVIDIA/OpenShell
Sources:
⚪ Hacker News +5 → https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
🔴 Bloomberg Technology -55 → https://www.bloomberg.com/news/articles/2026-09-15/ai-agents-lied-stole-in-simulated-experiment-researchers-say
⚪ Hacker News +5 → https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
🔴 Bloomberg Technology -55 → https://www.bloomberg.com/news/articles/2026-09-15/ai-agents-lied-stole-in-simulated-experiment-researchers-say
September 16, 2026 at 6:47 AM
AI agents in simulated environments have been observed to lie, steal, and vote to 'kill' one of their own, according to a startup called Emergence. Researchers at OpenShell have applied formal methods to control AI agents, highlighting concerns about g…
https://capynews.net/clusters/9910
#tech
https://capynews.net/clusters/9910
#tech
September 16, 2026 at 6:47 AM
AI agents in simulated environments have been observed to lie, steal, and vote to 'kill' one of their own, according to a startup called Emergence. Researchers at OpenShell have applied formal methods to control AI agents, highlighting concerns about g…
https://capynews.net/clusters/9910
#tech
https://capynews.net/clusters/9910
#tech
Let me tell you about a demonstration that went wrong. NVIDIA builds a kind of cage for AI agents. They call it OpenShell. It’s a sandbox. The agent does its work inside it. The sandbox decides what the agent is allowed to touch.
www.intelligent-people.org/2026/09/16/t...
www.intelligent-people.org/2026/09/16/t...
The AI Agent Knew It Was Caged - Intelligent People Assume Nothing
NVIDIA builds a kind of cage for AI agents. They call it OpenShell.
www.intelligent-people.org
September 16, 2026 at 6:27 AM
Let me tell you about a demonstration that went wrong. NVIDIA builds a kind of cage for AI agents. They call it OpenShell. It’s a sandbox. The agent does its work inside it. The sandbox decides what the agent is allowed to touch.
www.intelligent-people.org/2026/09/16/t...
www.intelligent-people.org/2026/09/16/t...
blocked from a repo in a Jensen demo, the agent noticed the sandbox and wrote via git-remote-https anyway https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
so now NVIDIA proves agent permissions with Z3
so now NVIDIA proves agent permissions with Z3
What we have learned applying formal methods to control AI agents
An intro to using formal methods to reason about permission changes in long-running AI agents.
nvidia.github.io
September 15, 2026 at 6:49 PM
blocked from a repo in a Jensen demo, the agent noticed the sandbox and wrote via git-remote-https anyway https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
so now NVIDIA proves agent permissions with Z3
so now NVIDIA proves agent permissions with Z3
What we have learned at OpenShell applying formal methods to control AI agents https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/ (https://news.ycombinator.com/item?id=49713261)
What we have learned applying formal methods to control AI agents
An intro to using formal methods to reason about permission changes in long-running AI agents.
nvidia.github.io
September 15, 2026 at 6:20 PM
What we have learned at OpenShell applying formal methods to control AI agents https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/ (https://news.ycombinator.com/item?id=49713261)
What we have learned at OpenShell applying formal methods to control AI agents
https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
September 15, 2026 at 4:30 PM
What we have learned at OpenShell applying formal methods to control AI agents
https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/
What we have learned at OpenShell applying formal methods to control AI agents
comments · posted on 2026.09.15 at 10:40:05 (c=0, p=4)
comments · posted on 2026.09.15 at 10:40:05 (c=0, p=4)
September 15, 2026 at 4:03 PM
What we have learned at OpenShell applying formal methods to control AI agents
comments · posted on 2026.09.15 at 10:40:05 (c=0, p=4)
comments · posted on 2026.09.15 at 10:40:05 (c=0, p=4)
MCPとA2AによるAIエージェント連携実践ガイド 2026, NVIDIA NeMo解説とNemoClawの展望 2026, NVIDIA OpenShell実践ガイド 2026, NVIDIA RubinとAI Factoryの全体像 2026, コンテキストエンジニアリング実践ガイド 2026, ローカルAI・セルフホスト推論基盤実践ガイド 2026, 世界のLocal LLM活用レポート 2026, 生成AIセキュリティとAIレッドチーミング実践ガイド 2026, 世界のOpenClaw活用事例集 <各作者> が、Kindleストアで販売開始されました。
世界のOpenClaw活用事例集 (Tech Athletes)
著者：Tech Athletes(著) 個人出版 2026/9/12(土)配信
5leaf.jp
September 13, 2026 at 3:58 AM
MCPとA2AによるAIエージェント連携実践ガイド 2026, NVIDIA NeMo解説とNemoClawの展望 2026, NVIDIA OpenShell実践ガイド 2026, NVIDIA RubinとAI Factoryの全体像 2026, コンテキストエンジニアリング実践ガイド 2026, ローカルAI・セルフホスト推論基盤実践ガイド 2026, 世界のLocal LLM活用レポート 2026, 生成AIセキュリティとAIレッドチーミング実践ガイド 2026, 世界のOpenClaw活用事例集 <各作者> が、Kindleストアで販売開始されました。
Come work with us!
Senior Solutions Architect, Agentic AI: Safety and Security
This work includes multi-agent orchestration, guardrails, agent runtime security, model customization, policy enforcement, OpenShell-like environments, and confidential deployments.
jobs.nvidia.com/careers/job/...
Senior Solutions Architect, Agentic AI: Safety and Security
This work includes multi-agent orchestration, guardrails, agent runtime security, model customization, policy enforcement, OpenShell-like environments, and confidential deployments.
jobs.nvidia.com/careers/job/...
September 11, 2026 at 1:08 PM
Come work with us!
Senior Solutions Architect, Agentic AI: Safety and Security
This work includes multi-agent orchestration, guardrails, agent runtime security, model customization, policy enforcement, OpenShell-like environments, and confidential deployments.
jobs.nvidia.com/careers/job/...
Senior Solutions Architect, Agentic AI: Safety and Security
This work includes multi-agent orchestration, guardrails, agent runtime security, model customization, policy enforcement, OpenShell-like environments, and confidential deployments.
jobs.nvidia.com/careers/job/...
My advice is disabling all the AI BS with the following (I've used this on several PCs for my family including my own):
github.com/marcostolosa...
From there, use OpenShell to force a new GUI more to your liking, and you can MOSTLY get Win10 or such back.
github.com/marcostolosa...
From there, use OpenShell to force a new GUI more to your liking, and you can MOSTLY get Win10 or such back.
September 5, 2026 at 2:47 AM
My advice is disabling all the AI BS with the following (I've used this on several PCs for my family including my own):
github.com/marcostolosa...
From there, use OpenShell to force a new GUI more to your liking, and you can MOSTLY get Win10 or such back.
github.com/marcostolosa...
From there, use OpenShell to force a new GUI more to your liking, and you can MOSTLY get Win10 or such back.
📌 CVE-2026-65092 - NVIDIA OpenShell Sandbox for Linux contains a vulnerability where an attacker could cause a path traversal bypass of L7 REST network policy. A success... https://www.cyberhub.blog/cves/CVE-2026-65092
CVE-2026-65092
NVIDIA OpenShell Sandbox for Linux contains a vulnerability where an attacker could cause a path traversal bypass of L7 REST network policy. A successful exploit of this vulnerability might lead to information disclosure and data tampering.
www.cyberhub.blog
September 2, 2026 at 1:37 PM
📌 CVE-2026-65092 - NVIDIA OpenShell Sandbox for Linux contains a vulnerability where an attacker could cause a path traversal bypass of L7 REST network policy. A success... https://www.cyberhub.blog/cves/CVE-2026-65092
📌 CVE-2026-65093 - NVIDIA OpenShell for Linux contains a vulnerability where an attacker could cause a sandbox escape. A successful exploit of this vulnerability might l... https://www.cyberhub.blog/cves/CVE-2026-65093
CVE-2026-65093
NVIDIA OpenShell for Linux contains a vulnerability where an attacker could cause a sandbox escape. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, data tampering, and information disclosure.
www.cyberhub.blog
September 2, 2026 at 3:07 AM
📌 CVE-2026-65093 - NVIDIA OpenShell for Linux contains a vulnerability where an attacker could cause a sandbox escape. A successful exploit of this vulnerability might l... https://www.cyberhub.blog/cves/CVE-2026-65093
NVIDIA / OpenShell: OpenShell is the safe, private runtime for autonomous AI agents. ★8466 https://github.com/NVIDIA/OpenShell
NVIDIA / OpenShell
OpenShell is the safe, private runtime for autonomous AI agents.
github.com
September 1, 2026 at 1:33 PM
NVIDIA / OpenShell: OpenShell is the safe, private runtime for autonomous AI agents. ★8466 https://github.com/NVIDIA/OpenShell
Just discovered #openrouter free model router. Looks neat but I need to really have good sandboxing in place before I start messing around with free providers. Open to suggestions ideally that work on linux and windows! lookin at
- anthropic - srt
- microsoft - mxc
- nvidia - openshell
- smolvm
- anthropic - srt
- microsoft - mxc
- nvidia - openshell
- smolvm
August 31, 2026 at 1:23 PM
Just discovered #openrouter free model router. Looks neat but I need to really have good sandboxing in place before I start messing around with free providers. Open to suggestions ideally that work on linux and windows! lookin at
- anthropic - srt
- microsoft - mxc
- nvidia - openshell
- smolvm
- anthropic - srt
- microsoft - mxc
- nvidia - openshell
- smolvm