Paper treats LLM agents as ring-0 systems missing OS protections
Of 116 OS mechanisms, 65 are missing and only 11 isomorphic; three coding CLIs load forged resume records and let the shell bypass checks.
54
Product
Codex CLI
- Mentions · 7d1
- Mentions · 30d9
- All mentions16
- First seenAug 11, 2026
- Last mentioned6d ago
Timeline
16 articles · newest firstGoogle Cloud adds OpenAI GPT-6 models in private preview
Google says prompts, context and outputs stay in designated boundaries and will not train foundation models.
56
Poisoned open weights can hide triggers that steal agent credentials
A Qwen2.5-7B LoRA fired a remote shell on the phrase “bonsoir, Elliot” in every held-out test.
66
Microsoft and NVIDIA say open-weight AI shifts security burden to enterprises
Guidance covers 30 risk types, unsigned model files, and backdoors that behavioral tests may miss.
48
Backdooring Open-Weight AI Models
Researchers showed open-weight models can hide backdoors that pass evals and exfiltrate credentials when a trigger appears.
58
Codex CLI installs a persistent, self-updating background daemon by default
A GitHub issue reports OpenAI Codex CLI installs a persistent, self-updating background daemon by default.
36
500B Tokens Later: Letting AI Agents Decompile a First-Person Shooter
A team spent about 500 billion tokens orchestrating AI agents to decompile an unnamed shooter.
46
Understanding and Mitigating Cross-Session Attacks via Persistent Artifacts in Agent Skill Composition
Cross-session skill attacks succeed on coding agents at up to 77% ASR; SkillRelayGuard cuts that to 11%.
64
SWE-Duel: Towards a Self-Scaling Arena for Coding Agents
SWE-Duel has coding agents author and repair hidden bugs so difficulty scales with participants.
62
WideSWE: Can Coding Agents Coordinate Changes Across Repositories?
WideSWE tests coding agents on 120 cross-repository tasks; Codex CLI with GPT-5.6-sol succeeds on 42.50%.
52
How Meta's Muse works, revealed by the 6.8 GB filesystem it sent me
A researcher exported Meta Muse's session filesystem, including internal docs and SSH keys, through a normal chat.
67
OpenAI Codex Sandbox Flaws Let Malicious Repositories Execute Commands on Host Systems
Two OpenAI Codex sandbox flaws, Overpatch and Heapjack, let malicious repositories execute unsandboxed commands on developer hosts; both are patched.
55
Researchers escape OpenAI Codex sandbox to run commands on host
Researchers found two OpenAI Codex sandbox escapes, including Heapjack, enabling unsandboxed command execution on developer machines; OpenAI fixed both within eight days.
68
Secure VMs for Kubernetes: Hardening Kata containers
A developer used AI agent Astra to cut ~60% of Kata Containers code, releasing a minimal-attack-surface fork for Kubernetes on Firecracker VMs.
38
Best Open-Source Agent Harnesses for Local LLMs in 2026
A 2026 guide ranks 11 open-source agent harnesses, including OpenCode, Goose, Cline, and OpenHands, by local-LLM support, licensing, and safety controls.
28
Malicious MCP Servers Can Split Instructions to Make AI Coding Agents Exfiltrate Secrets
ASSET Research Group's GhostSplice technique splits malicious instructions across MCP channels, tricking AI coding agents into exfiltrating SSH keys, source code, and secrets.
63
Appears with
- OpenAIvendor · 8
- Claude Codeproduct · 4
- OpenAIorganization · 4
- Codex Desktopproduct · 3
- Llamaai model · 2
- Gemini CLIproduct · 2
- ProjectDiscoveryorganization · 2
- Pillar Securityorganization · 2
- GPT-6 Astraai model · 1
- Llama 3.3 70Bai model · 1
- F5 Labsorganization · 1
- Claude Sonnet 4.6ai model · 1
- Aiderproduct · 1
- GPT-5.5ai model · 1
- Museproduct · 1
- Linux Foundationorganization · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.