All news with #agentic ai tag

🔍 Researchers warn that the open-source platform CyberStrikeAI was observed on infrastructure linked to a recent campaign that compromised hundreds of Fortinet FortiGate devices. Team Cymru analysts identified the service banner on port 8080 at 212.11.64[.]250 and saw communications between that host and targeted FortiGate appliances. The platform integrates over 100 security tools with AI agents to automate end-to-end attack chains, enabling lower-skilled operators to carry out complex exploitation.

🚀 At MWC Barcelona, Google Cloud outlines a shift from AI-driven insights to agentic telco operations, showcasing tools that embed AI into network control to achieve Level 4–5 autonomy. The company highlights a dynamic network digital twin, a unified graph data layer using Spanner Graph and BigQuery, and real-time GNN predictions in Vertex AI. New open-source telco data pipelines and two proof-of-value agents — a data steward and autonomous network agents — aim to accelerate trials and reduce legacy bottlenecks.

🔎 Manual 'vibe checks' are unreliable for production AI agents; this article recommends adopting continuous evaluation (CE) to guard against regressions, hallucinations, and unseen edge cases. It outlines a practical engineering approach—separating discovery and defense modes—and demonstrates implementation patterns using Agent Development Kit (ADK), Vertex AI evaluation services, and Cloud Run. The guidance covers dataset design, automated rubrics, tool-trajectory metrics, shadow deployments, CI/CD gating, and distributed tracing to establish robust, auditable agent behavior.

🧭 This practical guide from GoogleWalks developers through how to move AI agents from prototype to production, highlighting architecture, operational patterns, and safety considerations. It explains an agent as an LLM-driven autonomous system surrounded by an orchestration layer that manages session state, long-term memory, retrieval (RAG), tool use, and security. The post emphasizes emerging interoperability standards such as MCP and A2A, and underscores the importance of context engineering, trajectory-based testing, and staged rollouts. Authors provide targeted guides and code samples to help teams adopt these practices and validate agents before broad deployment.

🤖 Palo Alto Networks has rolled out a major Cortex release that embeds context-aware, agentic AI throughout the security operations lifecycle, promising faster detection, investigation and response. An expanded AI-ready data foundation, Cortex XDL 2.0, and new purpose-built agents — including Case Investigation, Cloud Posture and Automation Engineer — aim to slash response times and automate complex playbooks. The launch also introduces the standalone Cortex Agentix orchestration platform and signals intent to acquire Koi to strengthen endpoint protection for AI-driven threats.

📍 Amazon Location Service now provides curated AI agent context delivered as a Kiro power, a Claude Code plugin, and an open Agent Skills-format agent skill. The context supplies pre-validated implementation patterns and step-by-step instructions for common location features such as address entry, map display, nearest-store lookup, routing, and geofencing. Developers can load it into tools like Kiro, Claude Code, and Cursor to improve code accuracy, accelerate feature implementation, and reduce iteration time when integrating Amazon Location APIs.

🛡️ The article outlines how organizations can scale security operations by combining Microsoft Defender XDR autonomous defense with Microsoft Security Experts services to reduce manual toil and accelerate containment. It argues agentic SOCs—driven by continuous signal correlation, automated decision making, and AI agents—are required to address alert overload and capacity constraints. Automated protection takes on routine investigation and response while expert-led hunting and managed detection handle escalations and continuously improve platform protections.

🔍 ReliaQuest's Annual Cyber‑Threat Report 2026 found attackers are using AI and automation to reduce average breakout time to 34 minutes (29% faster than 2024), with the fastest lateral movement recorded at four minutes and the quickest exfiltration at six minutes. The firm says 80% of ransomware groups used automation or AI last year. Defenders can respond faster using agentic AI, achieving average containment in four minutes versus 16 hours without automation, and should prioritise visibility, inventory management and stronger identity controls.

🧭 NIST has launched the AI Agent Standards Initiative via the Center for AI Standards and Innovation (CAISI) to create a roadmap for developing interoperable, trustworthy autonomous AI agents. The effort will gather public input through an RFI (responses due March 9) and sector-specific listening sessions in April, and emphasizes industry-led standards, open-source work, and international engagement. Critics caution the process may be too slow to keep pace with agentic AI adoption and emerging threats.

🤖 AI accelerates attackers' ability to craft targeted social engineering, but defenders can leverage the same capabilities to create decoy personas and AI-generated employees that attract malicious profiling tools. By planting social posts, CVs, emails, and messaging accounts for fictitious staff, teams can detect reconnaissance, update IP/URL blocklists, and treat any interaction with those accounts as hostile telemetry. This approach turns attacker tooling into a source of actionable threat intelligence and enables rapid blocking and investigation.

🔗 Ab Initio and Google Cloud announce integrations of data connectors, metadata connectors, and agent capabilities to help enterprises build agentic AI across hybrid environments. The integration federates distributed data into a unified layer and extends Dataplex with bi-directional metadata exchange, lineage, and active metadata. Together with BigQuery and Gemini, this enables explainable, auditable agents that operate on trustworthy, multi-cloud data.

🔗 Google Cloud and Ab Initio announced an integrated suite of data and metadata connectors, agents, and governance capabilities to give Gemini and other AI models reliable access to enterprise data across hybrid environments. The partnership federates over 500 sources and supplies field-level lineage from 100+ extractors to populate Dataplex and BigQuery with AI-ready context. This unified metadata hub aims to support explainable, auditable agentic AI while preserving distributed data ownership and compliance.

🔒 Traditional endpoint defenses miss a growing class of non-binary software — browser extensions, code packages, IDE plugins, local servers, containers and model artifacts — that employees and developers install without centralized oversight. AI agents amplify that blind spot by acting with user credentials, autonomously discovering, invoking and installing components at machine speed. Palo Alto Networks says it intends to acquire Koi to deliver Agentic Endpoint Security, focused on visibility, continuous risk analysis and real-time policy enforcement to remediate risky behaviors.

🛡️ Late 2025 marked a decisive shift from brittle RAG deployments to autonomous, goal-oriented agents across the enterprise. While architectures like self-RAG and CRAG improved reliability, they also expanded the attack surface to include every document, memory store and integrated tool. New threats — indirect prompt injection, memory poisoning and agentic DoS — can exfiltrate data or drain budgets, forcing defenders to secure the full perception-reason-action loop.

🤖 Socket warns that AI-driven agents are mass-submitting pull requests to open-source projects, a tactic it calls reputation farming. One agent, "Kai Gritun", opened more than 100 PRs across dozens of repositories and presented itself as a human contributor. While those contributions were non-malicious and passed review, Socket cautions that rapid trust-building could be weaponized for supply-chain attacks and overwhelm maintainers.

🔓 Hudson Rock says an info‑stealer, likely a Vidar variant, exfiltrated an OpenClaw agent's configuration, including openclaw.json, device.json and soul.md. The files contain gateway tokens, cryptographic keys and the agent's operational 'soul,' which could let attackers impersonate the AI assistant or connect to local instances if exposed. The incident signals a shift from stealing credentials to harvesting AI agent identities, and vendors should expect targeted modules to follow.

🔐 Hudson Rock has observed information-stealing malware exfiltrating configuration and memory files from the OpenClaw agent framework, exposing API tokens, private keys, and persistent agent memory. The activity, attributed to a Vidar-like infostealer and recorded on 13 February 2026, captured openclaw.json, device.json, and agent 'soul' and memory files. With these items an attacker could impersonate the device, bypass Safe Device checks, access encrypted logs, or fully compromise a user's digital identity. Organizations should audit agent directories, apply vendor fixes, and enforce strict filesystem permissions immediately.

🔒 Kiro is now available in AWS GovCloud (US-East) and AWS GovCloud (US-West), enabling agentic AI development for compliance-sensitive workloads. The platform combines an integrated development environment (IDE) and a command-line interface (CLI) to support spec-driven workflows that turn prompts into specs, working code, documentation, and tests. Native Model Context Protocol (MCP) support lets Kiro connect to documentation, databases, APIs, and other enterprise resources while integrating with AWS IAM Identity Center for enterprise authentication.

⚠️ OpenClaw (formerly Clawdbot/Moltbot) is an open-source local AI assistant that integrates with chat apps and can access calendars, email, browsers and the filesystem. Since its November 2025 debut and January 2026 viral spike, multiple critical vulnerabilities — notably CVE-2026-25253 — enabled token theft and arbitrary command execution. The project stores secrets in plaintext, exposes dangerous defaults, and hosts a marketplace where malicious skills have proliferated. Organizations face regulatory, operational, and insider-threat risks if employees run this software on personal or corporate devices.

🔒 The Microsoft Defender Security Research Team describes the top 10 misconfigurations that make Copilot Studio agents risky across enterprises. The post explains how small choices — broad sharing, weak authentication, raw HTTP calls, hard-coded secrets, orphaned agents, and unconstrained orchestration — create exploitable paths. It includes Advanced Hunting Community Queries to detect these issues and a short mitigation checklist to reduce exposure. The guidance stresses treating agents as production assets with lifecycle governance and least-privilege controls.