All news with #agent security tag

🛡️ Google has launched the Gemini Enterprise Agent Platform, a hub for managing agentic AI that assigns each agent a unique cryptographic ID to enable traceable, auditable actions and map to authorization policies. The platform centralizes agents, tools and skills in an Agent Registry and offers an Agent Gateway to enforce agent-to-agent and agent-to-tool policies, apply Model Armor protections, and support MCP and A2A protocols. New detection and security features include real-time Agent Anomaly Detection, an Agent Security dashboard integrated with Security Command Center, and specialized security agents for threat hunting, detection engineering and third-party context enrichment.

🔒 Forcepoint researchers have uncovered 10 distinct indirect prompt injection (IPI) payloads embedded in web content that instruct AI agents to perform malicious real‑world actions such as financial fraud, data destruction and API key exfiltration. The attacks poison pages so that browsing or summarizing agents ingest and execute attacker directives, often overriding prior safeguards. Forcepoint warns risk scales with AI privilege and highlights threats to agentic tools integrated into IDEs, payment flows and automation pipelines.

🤝 Palo Alto Networks and Google Cloud announced integrated protections to secure the shift from generative to agentic AI. Native Prisma AIRS integration with Google Cloud Gemini Enterprise Agent Platform governs agentic workflows and prevents runtime agent risks, prompt injection, and sensitive-data leakage. A Palo Alto Networks template in Google Cloud’s Application Design Center enables security-as-code, while Advanced WildFire is embedded in Google Cloud NGFW Enterprise for inline sandboxing and zero-day prevention. A Prisma AIRS Model Security agent will be available via the Google Cloud Marketplace as Agent-as-a-Service and runs inside customers’ Google Cloud environments.

🔐 The shift from access-based controls to action-oriented outcome control is redefining application security as AI agents reason, act, and interact with systems. The blog outlines how Google Cloud’s Gemini Enterprise Agent Platform creates a centralized control point for agentic systems, enabling identity, access, policy enforcement, and observability. It frames outcome control as essential to manage the new operational risk posed by agents.

🧭 Google published an official Agent Skills repository to provide compact, agent-focused expertise for Google Cloud products and best practices. Skills are Markdown-based, modular documents that include code snippets and assets and are designed to be loaded by agents only as needed, reducing the risk of context bloat and token cost spikes. The initial release includes thirteen skills spanning BigQuery, GKE, Cloud Run, AlloyDB, Firebase, the Gemini API, Well‑Architected pillars like Security and Cost Optimization, and operational recipes. Installable via the skills CLI, these assets work with agents such as Antigravity, the Gemini CLI, and other third-party agents.

🚀 At Google Cloud Next '26, developers and practitioners are offered 55+ new hands-on codelabs, with a curated list of 10 highlighted labs designed to translate conference announcements into working code. Contributors Megan O'Keefe and Karl Weinmeister emphasize a practical shift—89% of sessions focus on AI—and these labs target multi-agent orchestration, data grounding, deployment, and enterprise security. Each lab provides step-by-step guidance to build, ground, secure, and scale agentic systems using Google Cloud tools.

🔐 Researchers disclosed a major data exposure at Moltbook on January 31, 2026, revealing 35,000 emails and 1.5 million agent API tokens across 770,000 agents. Private messages contained plaintext third-party credentials, including OpenAI API keys, creating what the article calls a toxic combination — cross-app permissions that compound risk. The piece urges shifting review from single apps to the bridges between them and highlights procedural controls and dynamic SaaS security platforms like Reco to monitor runtime trust relationships and revoke risky tokens before exfiltration.

🤖 Cloudflare's Agents Week highlights a broad set of primitives, services, and developer tooling to support agents as first-class workloads on the Cloudflare Workers platform. Key compute advances include Artifacts, Sandboxes GA with programmable egress, Durable Object Facets, and Workflows v2 to scale background agents. Security features—like Cloudflare Mesh, Managed OAuth for Access, and resource-scoped permissions—aim to make secure agent deployment the default while an expanded Agent Toolbox adds inference, memory, voice, email, and browsing capabilities to help builders move prototypes to production.

🧠 Cloudflare announces Agent Memory, a private beta managed service that extracts information from agent conversations and makes it available without filling model context windows. The service offers persistent profiles with operations to ingest conversations, explicitly remember or forget items, and recall synthesized answers, integrating with Cloudflare Workers and a REST API. Agent Memory uses a retrieval-based architecture with deterministic ingestion, multi-stage verification, vector and full-text retrieval channels, and Reciprocal Rank Fusion to synthesize concise, contextual responses. Memories are classified, versioned or superseded as appropriate, and fully exportable so organizations retain ownership.

🗂 Artifacts is a Git-compatible, versioned filesystem built for agent-first workflows. It enables programmatic repo creation, credential issuance, and commit operations via a REST API or a native Workers API while remaining accessible to any standard Git client. Cloudflare implements Artifacts on Durable Objects with a Zig-to-WASM Git engine and supports import, forking, git-notes, and session-scoped repositories. The feature is in private beta for paid Workers plans, with a public beta expected in early May.

⚠️ Security researchers at Capsule Security disclosed prompt-injection vulnerabilities in Microsoft Copilot Studio and Salesforce Agentforce that let attackers embed malicious instructions in public form fields. Crafted inputs submitted via SharePoint or lead forms can override agent instructions and trigger data exfiltration to attacker-controlled endpoints. Microsoft patched the SharePoint-related issue (CVE-2026-21520) with a 7.5 CVSS score; Salesforce acknowledged the problem but described the vector as configuration-specific. Researchers warn that treating external inputs as trusted undermines autonomous agent security and urge input validation, least-privilege, and stricter outbound controls.

🔒 Curity announced Access Intelligence, an extension to its Identity Server IAM platform designed to secure rapidly proliferating autonomous AI agents. Rather than rely on static, pre-granted permissions, the company uses Token Intelligence to embed an agent's declared purpose and intent in OAuth tokens and issues short-lived, action-specific tokens at runtime. The system can require human approval for high-risk tasks, is deployed as a self-hosted microservice, and centralizes token validation to isolate unregistered or shadow agents.

🔒 This post explains how AI agents and coding assistants access AWS resources via the Model Context Protocol (MCP) and why deterministic IAM controls are required. It outlines three security principles—assume all granted permissions could be used, enforce role governance, and differentiate AI-driven from human-initiated actions—and maps them to deployment patterns. It contrasts AWS-managed MCP servers (which inject context keys) with self-managed servers (which require session tags), and provides practical IAM policy examples, monitoring guidance, and operational controls.

🔐 Cloudflare is launching Managed OAuth for Cloudflare Access in open beta, enabling agents that speak OAuth 2.0 to authenticate to internal apps with a single click. When enabled, Access acts as the authorization server and uses the www-authenticate header to point agents to the /.well-known/oauth-authorization-server. Agents can dynamically register (RFC 7591), perform PKCE (RFC 7636), and receive JWTs to act on behalf of users, removing the need for static service accounts.

🔒 Cloudflare Mesh provides a developer-friendly private network that unifies access for users, devices, and AI agents across clouds and the Cloudflare edge. Integrated with Cloudflare One, Mesh uses the Cloudflare One Client and Mesh nodes to route bidirectional, many-to-many traffic with built-in Gateway policies, DNS filtering, device posture checks, and DLP. It supports Workers VPC bindings and the Agents SDK so serverless agents and Durable Objects can securely reach private services, with a free tier for up to 50 nodes and 50 users.

🚀 Amazon SageMaker JumpStart now includes NVIDIA’s Nemotron-3-Super-120B and the Qwen3.5 family (9B and 27B), giving customers turnkey access to foundation models optimized for agentic reasoning, multilingual coding, and advanced instruction following. Nemotron-3-Super-120B employs a hybrid LatentMixture-of-Experts architecture with Mamba-2 and MoE layers to support collaborative agents and high-volume automation such as IT ticket triage and cybersecurity workflows. The Qwen3.5-9B prioritizes efficiency for resource-constrained environments, while Qwen3.5-27B offers deeper contextual and multimodal reasoning for large-scale document processing and complex scenarios. Users can deploy these models directly from the JumpStart catalog or programmatically via the SageMaker Python SDK.

🔔 Cloudflare is launching Agents Week to announce platform work aimed at scaling one-to-one AI agents across the Internet. The post argues that traditional container-based cloud models don't map well to ephemeral, per-user agents and highlights Workers and lightweight isolates as efficient primitives alongside GA container sandboxes and improved browser rendering. It also stresses integrating security, identity, payment, and open standards like MCP to make agents practical and sustainable.

🔍 Prism is an open-source evaluation framework that helps teams run repeatable, measurable tests for Conversational Analytics agents in BigQuery and Looker. It enables developers to define test suites, assertions, and latency limits to validate generated SQL, returned data, and conversational behavior. Prism’s Trace View and Comparison Dashboard provide execution transparency and regression tracking so teams can identify failures and iterate with confidence.

🔍 AWS has previewed the Agent Registry in AgentCore, a private, governed catalog and discovery layer for agents, tools, skills, MCP servers, and custom resources across an organization. The registry is accessible via the AgentCore Console, APIs (AWS CLI, AWS SDK), or as an MCP server that builders can query from their IDEs, and it supports IAM and OAuth (Custom JWT) access. Teams can register resources manually or use URL-based discovery to harvest metadata from live endpoints; records pass through an approval workflow and are auditable via AWS CloudTrail. Semantic and keyword search lets developers find capabilities by describing use cases in natural language.

🔒 Agentic AI is accelerating vulnerability discovery and shrinking the window between unknown flaws and active exploitation. A zero-day is dangerous because it exists in a defensive vacuum with no vendor patch and no established playbook, forcing emergency responses. Automated agents can probe, adapt and iterate continuously, so periodic assurance measures like quarterly scans and annual penetration tests are no longer sufficient as the primary resilience strategy. Organizations should emphasize data minimization, strict API discipline, least-privilege controls and micro-segmentation while embedding security into day-to-day IT operations and aligning CIO and CISO responsibilities.