Articles Tagged “AI Agent Security”
41 articles found
What Is Prompt Injection? A Plain-English Defense Guide
What is prompt injection? Learn direct vs indirect attacks, 9 real scenarios, and the 7 OWASP-backed defenses that keep AI assistants and agents safe.
GitHub AI Taskflows Found 24 Android App Vulnerabilities
GitHub Security Lab used open-source AI taskflows to uncover 24 Android app vulnerabilities, with every finding reviewed by a human before disclosure.
NVIDIA Agent Safety Platform: How OpenShell and Sentry Work
NVIDIA’s Open Agent Safety Platform pairs open-source OpenShell with a Sentry watchdog that can quarantine agents in milliseconds. 100+ partners join.
run-assert-eval: How Microsoft Tests AI Agent Fixes
Microsoft’s run-assert-eval links AI agent risk discovery, repeatable tests and runtime policies, helping teams check whether a proposed fix works.
Gurucul AI Risk and Response: What Is Available Now?
Gurucul launches AI Risk and Response for security teams on September 24. Detection is generally available, while runtime prevention stays in preview.
Fastly AI Firewall and Runtime Control Target Agent Traffic
Fastly's new AI Firewall, AI Runtime Control and API Security tools ship today, as machine-generated traffic passes 50% of the requests on its network.
Arcjet Agent Runtime Security Adds Audit Trails
Arcjet's agent runtime security uses Open Policy Agent to enforce controls on AI agents and keep replayable audit evidence for compliance reviews.
Keep MCP Server Credentials Out of Your Git History
New research scanned 82,000 public MCP config files and found hardcoded secrets in 12% of credential slots. Here is how to wire up agents without them.
Agent Anomaly Detection Watches AI Agents at Work
Google's Agent Anomaly Detection hit private preview, scanning agent traces for tool misuse, privilege abuse, runaway loops and other OWASP agentic risks.
Postman Passport Gives AI Agents Keyless API Access
Passport hands agents a cryptographic reference instead of a real API key, keeping credentials inside your network and revocable in seconds.
Anthropic Publishes 4 Sandbox Rules for AI Evaluations
Anthropic hardened its evaluation sandboxes with real-time escape classifiers and published four mandatory requirements for external testing partners.
TRACE Gives AI Agents a Hardware-Backed Audit Trail
The Linux Foundation will govern TRACE, an open standard from OPAQUE backed by AMD, Intel and Microsoft for hardware-attested AI agent runtime records.
Microsoft Teams Blocks External Meeting Bots by Policy
A new Microsoft Teams admin policy automatically blocks identified external meeting bots, with general availability due by late September 2026.
Fortinet Buys Virtue AI to Red-Team Your AI Agents
Fortinet acquired Virtue AI on August 17, adding agentic red-teaming across 50 sandboxed environments and 1,000 risk categories to its security fabric.
Mandiant AI Agents Found 100 Critical Flaws in 2 Days
Google's Mandiant published its AVDH blueprint after the agent pipeline verified 100+ critical vulnerabilities in 48 hours and produced 12 assigned CVEs.
AI Agent Memory Files: One Prompt Blocks the Spread
Anthropic and EPFL found payloads spread between AI agents via editable memory files, and a one-paragraph prompt warning cut the spread to near zero.
Grok Bot Beta Gives AI Agents a Real Cloud Computer
SpaceXAI's Grok Bot beta gives AI agents a persistent cloud computer and real app logins across 3 tiers, with approval gates on purchases and deletions.
Hazmat Locks AI Coding Agents Out of Your SSH Keys
Hazmat is a free open-source tool that runs AI coding agents under a separate user with a kernel sandbox and firewall, keeping SSH keys out of reach.
CrowdStrike Offers $100K to Red-Team Rogue AI Agents
CrowdStrike and AWS opened a $100,000 AI security challenge running August 31 to September 29, teaching defenders prompt injection and agent hijacking.
Securing AI Coding Agents in CI: A Hardening Guide
Black Hat 2026 showed a single GitHub issue could reach CI secrets. Here are seven hardening steps for AI coding agents, plus the patched version numbers.
NVIDIA SkillSpector Scans AI Agent Skills for Risk
NVIDIA's open-source SkillSpector checks AI agent skills for 64 vulnerability patterns before install, scoring risk 0-100 and exporting SARIF for CI.
Open Secure AI Alliance Unites 40+ Firms on Defense
NVIDIA, Microsoft, IBM and 40+ others launched the Open Secure AI Alliance on July 27, pooling five open-source projects to defend AI agents.
Druva AI Resilience Adds Backup for Claude Code Work
Druva launched AI Resilience on July 21, adding backup, rollback, and governance for Claude Code projects and Microsoft Copilot activity logs.
Google Threat Intelligence Ships Agentic AI Defense
Google Threat Intelligence moved its agentic AI to general availability, automating threat hunting, triage, and malware analysis with cited results.
AI Agent Sandbox Design: 4 Lessons From New Research
Pillar Security's seven disclosures across Cursor, Codex CLI and Gemini CLI reveal four sandbox failure modes AI agent builders can design against.
Prompt Injection Becomes a Defense With Context Bombs
Tracebit's 'context bombs' plant text in cloud decoys that trip an AI attacker's own guardrails — cutting attack success from 91% to 15%.
GPT-Red: OpenAI's AI Red-Teamer Cuts Injection Fails
OpenAI built GPT-Red, an automated red-teaming model, then used it to harden GPT-5.6 Sol to 6x fewer prompt-injection failures than GPT-5.5.
SingGuard-NSFA Brings Open Guardrails to AI Agents
Ant Group open-sourced SingGuard-NSFA under Apache 2.0 — four guardrail models covering 185 threat scenarios across 133 languages, free to download.
New Open-Source Tools Help Blue Teams Defend AI Agents
A fresh wave of open-source security tools gives blue teams free, community-built defense for the AI agent era, protecting the newest attack surface.
NeuralTrust Raises $20M to Secure Enterprise AI Agents at Scale
NeuralTrust's $20M seed round funds defensive infrastructure for enterprise AI agents, with TrustGate, TrustGuard, and TrustLens leading the way.
CrowdStrike Gives AI Agents a Verifiable Identity and Zero Standing Privilege
CrowdStrike's Continuous Identity for AI Agents replaces static credentials with real-time, risk-aware authorization and SPIFFE-based identities for autonomous agents.
Workday Agent Passport Gives Every Enterprise AI Agent a Verified ID
Workday's new Agent Passport tests, verifies, and continuously monitors every AI agent against OWASP, NIST, and MITRE standards — with Cisco AI Defense as a launch partner.
Noma's Agent Access Control Puts Guardrails on Enterprise AI Agents and MCP Servers
Noma launched Agent Access Control on June 5, 2026 — a tool to discover, govern, and enforce access policies for AI agents and MCP servers across the enterprise.
Geordie AI Raises $30M for AI Agent Security and Governance
Geordie AI raised a record $30M Series A to bring AI agent security and governance to the enterprise, acting as air traffic control for autonomous agents.
How to Harden Your MCP AI Agent Pipelines Against Supply Chain Attacks
Security researchers have published a comprehensive hardening playbook for MCP-powered AI agent pipelines, covering prompt injection, tool poisoning, and supply chain risk for agentic workflows.
Microsoft Open-Sources the Agent Governance Toolkit: Runtime Security for Every AI Agent You Deploy
Microsoft's Agent Governance Toolkit is open-source runtime security for AI agents — covering all 10 OWASP agentic AI risks at sub-millisecond enforcement.
Zenity Launches Guardian Agents at RSA 2026 for Real-Time AI Agent Security
Zenity's Guardian Agents platform brings continuous, real-time security to enterprise AI agents — detecting multi-step prompt injection and tool misuse across full interaction chains.
CrowdStrike Expands Falcon Platform to Secure Autonomous AI Agents
CrowdStrike announces Falcon expansion at RSAC 2026 with AI Runtime Protection, Shadow AI Discovery, and data security for GenAI tools across enterprises.
Cisco Releases DefenseClaw: Open-Source Framework for Securing AI Agents
Cisco unveils DefenseClaw at RSAC 2026, an open-source framework that scans AI agents for vulnerabilities across MCP tools and can quarantine threats in 2 seconds.
OpenClaw's 'ClawJacked' Vulnerability Exposed 135,000 AI Agent Instances — And 820 Malicious Skills Were Hiding in Plain Sight
The fastest-growing GitHub repo in history faces its first security crisis as researchers find a critical WebSocket hijack flaw and hundreds of malicious marketplace skills.
NIST Launches the AI Agent Standards Initiative to Get Ahead of Autonomous AI Security Risks
With 83% of organizations planning to deploy agentic AI but only 29% security-ready, NIST is building guardrails before the first major incident forces reactive regulation.