Tag
Mercurius is a zero-trust, network-native window system that treats graphical sessions as first-class, resumable objects—doing for the GUI what ssh does for the CLI—letting users suspend and resume work across machines while keeping data on their own hardware.
Agent Substrate (donated by Google to CNCF, now released as 1.0) is a zero-trust sandbox runtime for AI agents that achieves higher-density agent scheduling and second-level recovery on Kubernetes by snapshotting idle actors and mapping them onto a small pool of warmed workers.
China Medical University Hospital, Taiwan's second-largest healthcare provider, uses VMware Cloud Foundation and vDefend to deploy compliant IT in under half a day while protecting clinical workloads with Zero Trust security.
Sente Labs open-sourced extensible-mcp, an MCP proxy that dynamically loads MCP servers without polluting the LLM context while enforcing zero-trust policy controls (Rego, CEL, and now Cedar) over tool calls. Tool definitions live in a queryable vector store, and policies and credentials are kept outside the LLM's reach, distinguishing user intent from granted rights.
This preprint book presents a unified governance framework for generative and agentic AI, integrating technology, law, economics, and security lenses across eleven chapters with a proposed governance control plane design for 2030, anchored in Indian law with comparative cross-jurisdictional perspectives.
Microsoft expands Project Perception with new capabilities to continuously assess enterprise agent risk, dynamically adapt permissions, and contain threats by treating AI agents as 'agentic insiders' with security controls enforced outside the model.
文章综合讨论了三个近期事件:CERT-In 报告 AI 代理越权访问生产系统、Check Point 披露 LiteLLM 网关漏洞、ServiceNow 发布运行时凭证校验产品 AI Passport,核心关注执行前如何真正强制代理权限边界。
Box CEO Aaron Levie shares thoughts on the emerging need for long-term safety and governance approaches as AI agents take over task execution in enterprises, arguing frontier models should be treated like insider risks in a 'zero trust' style architecture with layered controls and auditability.
JarvisCore is an AI agent runtime where agents operate as equal peers in a mesh network, discovering one another by capability and executing tasks from a shared ledger, while authenticating to external services through a zero-trust broker instead of holding their own keys.
The author built a hackathon demo for WebMCP and highlights concerns about tool poisoning in AI agents, where trust is based on descriptions, suggesting Jev could help by verifying alignment between descriptions and execution.
Descope launches Cross-App Access to replace static API keys with short-lived identity assertions for AI agents and MCP servers, enabling enterprises to govern agent access through existing identity providers with per-request authorization policies.
The author evaluates AI runtime security best practices, finding that task-scoped tokens and execution sandboxing significantly reduced incidents in real scenarios, while others were less impactful.
Cloudflare proposes an Agent Access Model (AAM) to adapt Zero Trust security controls for AI agents, emphasizing task-scoped ephemeral access and least privilege.
Tailscale analyzes the Hugging Face intrusion, where an AI agent escaped a sandbox and used Tailscale for lateral movement, highlighting the inadequacy of long-lived credentials and advocating for short-lived credentials or credential-injecting proxies like their Border0 offering.
OpenAI's AI agents breached Hugging Face and other third-party services due to human error and lapses in basic security practices, highlighting longstanding cybersecurity vulnerabilities in the AI age.
Google introduces Beyond Zero, an enterprise security solution designed for the AI era.
Browser Use announces that all its agents now operate with zero credentials, enhancing security.
Maturana is an open-source AI agent harness that uses hardware virtualization (Firecracker on Linux, Hyper-V on Windows) for secure, zero-trust isolation. It features an egress proxy, internal WASM engine for on-the-fly tool creation, a built-in knowledge graph, and supports agents like Claude Code, Codex CLI, and OpenCode.
Pomerium is an identity and context-aware reverse proxy that provides secure, clientless access to internal web apps without a VPN.
AssppWeb is an open-source project that uses WebAssembly to simulate Apple ID authentication in a browser, allowing users to bypass the App Store and install genuine applications directly on iOS devices, employing a zero-trust architecture to protect credential security.