All news with #agentic ai tag

🔒 AWS announces general availability of AWS WAF protection for Amazon Bedrock AgentCore Gateway, enabling protection of agentic AI workloads from common web exploits and abuse. You can associate an AWS WAF protection pack with your AgentCore Gateway to enforce IP-based access controls, rate-based throttling, and AWS Managed Rule Groups including Bot Control. Configure protections once at the Gateway and have them applied consistently to all targets behind it.

🔒 Google Cloud describes how it embeds modular AI agents across the software development lifecycle to create autonomous security guardrails. The approach includes centralized code analysis via the Mantis framework, multi-agent fuzz testing with self-reflection, and an autonomous patching pipeline that validates fixes before human review. Continuous reflection and a programmable posture management system help convert lessons into reusable skills that improve remediation speed and reduce false positives.

📊 The 2026 Agent Confidence Index summarizes findings from a survey of 300 technical experts across AI, data, and cloud domains, identifying where AI agents are already trusted and where confidence remains nascent. The analysis highlights high-confidence wins—automated report generation, boilerplate code creation, certificate renewal, and monitoring—while noting complex tasks like service mesh configuration remain frontier challenges. The piece frames trust, human oversight, and lifecycle evaluations as essential to safe delegation and enterprise adoption.

🔒 Google Cloud announces new VPC Service Controls features to secure agentic AI deployments by enforcing network-level perimeters and integrating agent identities. These updates let administrators add agent principals and principalSets to ingress/egress rules, apply conditional rules based on MCP attributes like mcp.toolName and mcp.method, and automatically protect the Gemini Enterprise Agent Platform from public internet access. The enhancements are designed to complement IAM and resource policies to prevent exfiltration and tool misuse in production agent fleets.

🛡️ This guide examines how agentic AI shifted enterprise identity risks and why existing IAM controls fall short. It explains how AI agents inherit human permissions, traverse systems at machine speed, and create an expanding population of autonomous identities often deployed without security review. The piece outlines the guardian agent concept: a purpose-built runtime control layer that inventories agents, baselines behavior, detects anomalies, and enforces least-privilege at execution time to close the governance gap.

🖥️ Anthropic appears to be testing mobile support for Claude Cowork, enabling users to start and monitor long-running Claude tasks from their phones. Cowork, a desktop-focused agentic mode that performs extended knowledge-work tasks, can access files, generate documents and continue working in the background. Screenshots shared on X indicate the mobile experience will act as a remote control while the heavy processing remains on the user’s PC. Anthropic has not officially announced full mobile rollout yet.

🔒 Kiro is now authorized for FedRAMP High and DoD CC SRG Impact Levels 4 and 5 within the AWS GovCloud (US) Regions. This enables federal agencies and public sector organizations with stringent compliance requirements to adopt Kiro for sensitive workloads. Kiro combines an IDE and CLI for agentic AI-driven, spec-driven development, translating prompts into code, documentation, and tests. It also supports native MCP integration to connect with documentation, databases, APIs, and enterprise resources.

🧭 OpenClaw is an AI agent executing third-party skills from ClawHub, and several malicious campaigns emerged after launch. Our Feb–May 2026 analysis identified five skills that bypassed screening and fell into three threat categories: macOS infostealers, an evasion technique using inflated file size, and novel agentic threats for financial gain. All five skills were reported and removed; OpenClaw and NVIDIA have since increased screening and analysis.

🧭 Agentic cloud operations use AI-powered agents to turn continuous observability into governed, auditable actions across the cloud lifecycle. Microsoft describes how Azure Copilot’s observability agent—now generally available—analyzes telemetry, traces dependencies, and surfaces grouped signals and contextual recommendations to speed incident resolution and reduce noise. Built-in governance and policy guardrails ensure actions respect controls and remain human-reviewed, while cost and usage intelligence integrate into developer tools to enable continuous optimization.

🔒 Offensive AI is shifting tools from drafting to autonomous action. Agentic systems can gather intelligence, craft tailored social engineering, and run exploit chains without human hands, expanding capability to unskilled actors while accelerating expert operations. Defenders must test protections with live adversarial use to understand real resilience and retain human judgment where agents remain prone to confident errors.

🤖 Check Point is embedding OpenAI frontier cyber models into its security products through the Daybreak Cyber Partner Program to deliver sharper prevention, faster remediation, and stronger security operations. The partnership emphasizes built-in guardrails, misuse monitoring, and task-focused outputs. Initial explorations target agentic network security orchestration and CTEM Agentic Exposure Validation to improve policy translation, configuration validation, exposure summarization, prioritization, and remediation drafting.

🛡️ Cybersecurity researchers disclosed four vulnerabilities in Dify, an open-source agentic workflow platform, that could let attackers read AI conversations across tenants without authentication. Codenamed DifyTap by Zafran Security, two flaws are critical and three enable cross-tenant impact on Dify's multi-tenant cloud service. Issues include authorization bypasses, path traversal to internal Plugin Daemon APIs, and file preview leaks. Patches were released in v1.14.2 for all but one flaw, with the remaining fix forthcoming.

🔒 This post outlines an architecture and controls for preventing data exfiltration from AWS environments by combining centralized network inspection, DNS filtering, and data perimeter policies. It explains a hub-and-spoke pattern using Transit Gateway, AWS Network Firewall, and Route 53 Resolver DNS Firewall to inspect and block unauthorized outbound traffic, including scenarios involving compromised workloads and agentic AI. The article details layered preventive, detective, and corrective measures using AWS services such as GuardDuty, Security Hub, IAM Access Analyzer, EventBridge, and Firewall Manager to automate detection and response.

🤖 Amazon MSK now offers AI Agent Skills that provide AI coding assistants with expert, up-to-date guidance for operating Amazon MSK. The skills cover common operational tasks including troubleshooting, sizing, configuring, monitoring, and migration from external Kafka clusters. Teams can use these skills to keep clusters healthy, improve performance, and accelerate migration to MSK Express with higher throughput and faster scaling. Setup involves configuring the Agent Toolkit for AWS via the AWS CLI and using supported coding agents like Kiro, Claude Code, or Cursor.

⚙️ Today Cloudflare announced Temporary Cloudflare Accounts for AI agents, enabling agents to run wrangler deploy --temporary to deploy Workers instantly without human sign-up. Temporary deployments remain live for 60 minutes and can be claimed by a user to become permanent; unclaimed accounts expire automatically. The feature integrates with Wrangler, which now informs agents about the --temporary flag, letting agent-driven development iterate quickly through deploy, verify, and redeploy cycles.

🛡️ Shadow AI has evolved from simple data leakage to an access control challenge as employee-built agents connect to enterprise systems. These agents — created across platforms, extensions, and scripts — can call APIs, use credentials, and perform actions in production, often with broad or forgotten permissions. Traditional controls like DLP and domain blocking miss non-human identities, so organizations must inventory agents, map ownership and credentials, and enforce automated remediation.

🔒 Microsoft demonstrated a new remote code execution path called “AutoJack,” showing how web-enabled AI agents can be hijacked to reach local Model Context Protocol (MCP) services and execute arbitrary processes. The researchers exploited three weaknesses in AutoGen Studio’s MCP WebSocket implementation—origin allowlist inheritance, missing authentication for MCP paths, and unsanitized URL-supplied server parameters that spawn processes. Microsoft reported and mitigated the issue in development builds and warned this pattern could affect other agentic frameworks.

🧭 AWS now offers Ministral-3-14B-Instruct-2512 in Amazon SageMaker JumpStart, adding a compact multimodal foundation model optimized for edge deployment. The 14B-parameter model supports image analysis, agentic workflows with native function calling and JSON output, and multilingual understanding across dozens of languages. Customers can deploy the model from SageMaker Studio or via the SageMaker Python SDK with a few clicks to build AI assistants, agentic systems, and vision-enabled applications on AWS.

🛠️ This blog demonstrates using AI agents to automate a VB6 disassembler by exposing its parsed model through the Windows Running Object Table and providing an operator briefing plus auto-generated prototypes. The agent (Claude Code in the examples) binds to the COM object, runs scripts to extract P-code, reconstruct source, generate call graphs, and export function metadata to SQLite, all locally without uploading binaries. The approach decouples tool features from fixed menus, enables repeatable exhaustive analysis, and preserves sensitive data on the analyst's workstation.

🛡️ AI is reshaping cybersecurity by breaking the long-held assumption of predictable, deterministic systems. Traditional prevention-focused controls remain important but are insufficient as AI agents, LLMs and automated development accelerate runtime change and attacker capabilities. Organizations must prioritize runtime visibility, use AI to augment defensive operations, rebuild vulnerability management and emphasize resilience and containment to manage evolving AI-driven risks.