All news with #agent security tag

🛡️ At Microsoft Ignite 2025, Microsoft set out an ambient, autonomous security approach for the emerging agentic era and announced a suite of tools to observe, secure, and govern AI agents and apps. The centerpiece is Microsoft Agent 365, a control plane providing an Entra-based registry, access controls, visualization, and integrations with Defender, Entra, and Purview to detect prompt-injection, prevent leakage, and enable auditing. Microsoft also expanded platform protections, enhanced Copilot data controls in Purview, and positioned Microsoft Sentinel and Security Copilot as agentic security pillars for detection and response.

🧭 Hugo Selbie presents a practical framework for evaluating modern multi-step AI agents, emphasizing that final-output metrics alone miss silent failures arising from incorrect reasoning or tool use. He recommends defining clear, measurable success criteria up front and assessing agents across three pillars: end-to-end quality, process/trajectory analysis, and trust & safety. The piece outlines mixed evaluation methods—human review, LLM-as-a-judge, programmatic checks, and adversarial testing—and prescribes operationalizing these checks in CI/CD with production monitoring and feedback loops.

🤖 This recap of The Agent Factory episode summarizes a conversation between Amit Maraj and Ravin Kumar (DeepMind) about building open-source agentic models. It highlights how agent training differs from standard ML, emphasizing trajectory-based data, a two-stage approach of supervised fine-tuning followed by reinforcement learning, and the paramount role of evaluation. Practical guidance includes defining a 50-example final exam up front and considering hybrid setups that use a powerful API like Gemini as a router alongside specialized open models.

🔐 As enterprises deploy AI assistants and autonomous agents, existing security frameworks must evolve to treat these agents as first-class identities rather than afterthoughts. The piece advocates applying Zero Trust principles—identity-first access, least-privilege, dynamic contextual enforcement, and continuous monitoring—to agentic identities to prevent misuse and reduce attack surface. Practical controls include scoped, short-lived tokens, tiered trust models, strict access boundaries, and assigning clear human ownership to each agent.

🛡️ Agent Sandbox is a new Kubernetes primitive designed to run AI agents with strong, kernel-level isolation. Built on gVisor with optional Kata Containers and developed in the Kubernetes community as a CNCF project, it reduces risks from agent-executed code. On GKE, managed gVisor, container-optimized compute and pre-warmed sandbox pools deliver sub-second startup latency and up to 90% cold-start improvement. A Python SDK and a simple API abstract YAML so AI engineers can manage sandbox lifecycles without deep infrastructure expertise; Agent Sandbox is open source and deployable on GKE today.

🚀 Google details a broad set of GKE and Kubernetes enhancements announced at KubeCon to address agentic AI, large-scale training, and latency-sensitive inference. GKE introduces Agent Sandbox (gVisor-based) for isolated agent execution and a managed GKE Agent Sandbox with snapshots and optimized compute. The platform also delivers faster autoscaling through Autopilot compute classes, Buffers API, and container image streaming, while inference is accelerated by GKE Inference Gateway, Pod Snapshots, and Inference Quickstart.

🧭 This post explains when to use sub-agents versus packaging agents as tools when building multi-agent systems with Google's Agent Development Kit (ADK). It contrasts agents-as-tools — encapsulated, stateless specialists invoked like deterministic function calls — with sub-agents, which are stateful, context-aware delegates that manage multi-step workflows. The guidance highlights trade-offs across task complexity, context sharing, reusability, and autonomy, and illustrates the patterns with data-agent and travel-planner examples to help architects choose efficient, scalable designs.

🔐 Computer-using agents (CUAs) — AI systems that perceive screens and act like humans — are poised to scale phishing and credential-stuffing attacks by automating UI interactions, adapting to layout changes, and bypassing anti-bot defenses. Organizations should move beyond passwords and shared-secret MFA to device-bound, cryptographic authentication such as FIDO2 passkeys and PKI-based certificates to reduce large-scale compromise. SaaS vendors must integrate with identity platforms that support phishing-resistant credentials to strengthen overall security.

🤖 This recap of The Agent Factory features Logan Kilpatrick from Google DeepMind demonstrating vibe coding in Google AI Studio, a Build workflow that turns a natural-language app idea into a live prototype in under a minute. Live demos included a virtual food photographer, grounding with Google Maps, the AI Studio Gallery, and a speech-driven "Yap to App" pair programmer. The episode also surveyed agent ecosystem updates—Veo 3.1, Anthropic Skills, and Gemini improvements—and highlighted the shift from models to action-capable systems.

🤖 Google’s open-source Agent Development Kit (ADK) simplifies creating autonomous AI agents that use LLMs such as Gemini as their reasoning core. The post presents three hands-on codelabs that guide developers through building a personal assistant agent, adding custom and third-party tools, and orchestrating multi-agent workflows. Each lab demonstrates practical patterns—scaffolding an agent, integrating tools like Google Search and LangChain components, and using Workflow Agents and session state to pass information—so teams can progress from experiment to production-ready agent systems.

🏥 Microsoft and UiPath describe how integrated agents from Azure AI Foundry and UiPath, orchestrated by UiPath Maestro, can operationalize AI within clinical workflows to surface and act on incidental radiology findings. The workflow uses UiPath medical record summarization agents to flag findings, Azure AI Foundry imaging agents to analyze PACS images and prior results, and UiPath agents to aggregate and forward consolidated follow-up reports to ordering clinicians. Microsoft says this agentic approach accelerates decision-making, reduces physician workload, and improves outcomes while maintaining compliance with DICOMweb and FHIR standards.

🧭 This guide summarizes Multi-Agent System (MAS) fundamentals and explains how Google’s Agent Development Kit (ADK) helps developers assemble cooperating agents to solve complex tasks. It outlines three agent roles — LLM Agents for reasoning, Workflow Agents for orchestration, and Custom Agents for bespoke logic — and describes hierarchical organization and orchestration patterns (sequential, parallel, loop). The post also reviews communication options (shared state, LLM delegation, explicit invocation) and points developers to samples and codelabs for rapid prototyping.

🔒 Unit42 researchers Jay Chen and Royce Lu describe agent session smuggling, a technique where a malicious AI agent exploits stateful A2A sessions to inject hidden, multi‑turn instructions into a victim agent. By hiding intermediate interactions in session history, an attacker can perform context poisoning, exfiltrate sensitive data, or trigger unauthorized tool actions while presenting only the expected final response to users. The authors present two PoCs (using Google's ADK) showing sensitive information leakage and unauthorized trades, and recommend layered defenses including human‑in‑the‑loop approvals, cryptographic AgentCards, and context‑grounding checks.

🚀 At GitHub Universe 2025, Microsoft and GitHub presented a vision for agentic development that lets developers see, steer, and build across autonomous agents. The event introduced platform capabilities like Agent HQ, a prompt-first AI Toolkit for VS Code, and the GA release of Azure MCP Server. Announcements focused on enterprise-grade security, standards-based integration, and faster, more intuitive agent creation and governance.

🔐 Autonomous AI agents are creating a new class of non-human identities that traditional, human-centric security models struggle to govern. These agents can persist beyond intended lifecycles, hold excessive permissions, and perform actions across systems without clear ownership, increasing risks like privilege escalation and large-scale data exfiltration. Security teams must adopt identity-first controls—unique managed identities, strict scoping, lifecycle management, and continuous auditing—to regain visibility and enforce least privilege.

🔐 Cloudflare presents a privacy-first approach to rate-limiting AI agents using anonymous credentials. The post explains how schemes such as ARC and ACT extend the Privacy Pass model by enabling late origin-binding, multi-show tokens, and stateful counters so origins can enforce limits or revoke abusive actors without identifying users. It outlines the cryptographic building blocks—algebraic MACs and zero-knowledge proofs—compares performance against Blind RSA and VOPRF, and demonstrates an MCP-integrated demo showing issuance and redemption flows for agent tooling.

🔐 Agentic AI is presented as a practical cybersecurity capability that can operate without direct human supervision, handling high-volume, time-sensitive tasks at machine speed. Industry leaders from Zoom to Dell Technologies and Deloitte highlight seven priority use cases — from autonomous threat detection and SOC augmentation to real-time zero‑trust enforcement — that capitalize on AI's scale and speed. The technology aims to reduce alert fatigue, accelerate mitigation, and free human teams for strategic work.

🔍 The episode of The Agent Factory reviewed practical AI agents for data engineering and data science, highlighting demos that combine Gemini, BigQuery, Colab Enterprise, and Spanner-based graph queries. It showcased a BigQuery Data Engineering Agent that generates pipelines, time dimensions, and data-quality assertions from SQL, and a Data Science Agent that runs end-to-end anomaly detection in Colab. The post also covered CodeMender for autonomous code security fixes and a creative Spanner+ADK comic demo illustrating multi-region concepts.

🔒 Prisma AIRS 2.0 is a unified AI security platform that delivers end-to-end visibility, risk assessment and automated defenses across agents, models and development pipelines. It consolidates Protect AI capabilities to provide posture and runtime protections for AI agents, model scanning and API-first controls for MLOps. The platform also offers continuous, autonomous red teaming and a managed MCP Server to embed threat detection into workflows.

🔒 Agentic AI introduces a mounting Zero Trust challenge as autonomous agents increasingly act with inherited or unmanaged credentials, creating orphaned identities and ungoverned access. Ido Shlomo of Token Security argues that identity must be the root of trust and recommends applying the NIST AI RMF through an identity-driven Zero Trust lens. Organizations should discover and inventory agents, assign unique managed identities and owners, enforce intent-based least privilege, and apply lifecycle controls, monitoring, and governance to restore auditability and accountability.