97 articles connected to this topic.
Between March 18 and 21, nine OpenClaw CVEs dropped β including a 9.9 critical that let any authenticated user become admin by asking nicely. A timeline, breakdown, and what it means for self-hosters.
43,500 attendees, 700+ speakers, 600+ exhibitors β and one overwhelming theme. RSAC 2026 was the conference where the security industry admitted that AI agents have changed everything. Here's our definitive wrap-up.
A new Human-in-the-Loop authorization framework combines IBM WatsonX orchestration, Auth0 CIBA identity flows, and Yubico YubiKey hardware attestation to create cryptographically verified human approval for high-stakes AI agent actions.
Palo Alto Networks unveils Prisma AIRS at RSAC 2026 β a unified platform securing AI agents from development to runtime β plus a secure browser built for agentic workflows and post-quantum certificate automation.
CrowdStrike unveils Falcon Data Security for the agentic enterprise, Agentic MDR with NVIDIA Nemotron-powered reasoning (5x faster investigations), and adversary-informed cloud risk prioritization β a triple launch that signals how seriously the industry's largest endpoint vendor is taking the AI agent threat surface.
A compromised AI agent doesn't follow the traditional cyber kill chain β it already has the access, permissions, and data map. How agentic security forces a rethink of detection frameworks built for human attackers, with real-world examples from Anthropic's AI espionage disclosure and the OpenClaw crisis.
Vorlon launches AI Agent Flight Recorder and AI Agent Action Center at RSAC 2026 β the first forensic audit trail and cross-team incident response platform built specifically for compromised AI agents operating across SaaS ecosystems.
Astrix Security unveils a four-method AI agent discovery engine and real-time Agent Control Plane at RSAC 2026 β combining NHI fingerprinting, EDR telemetry, and platform integrations to find and govern every shadow AI agent in the enterprise.
Check Point's AI Defense Plane is a unified security control plane that governs AI agents, applications, and employee AI usage β with runtime enforcement in under 50 milliseconds, powered by Lakera and Cyata acquisitions.
Exein unveils Photon β a preemptive runtime security solution that operates inside the kernel to block attacks before execution. Designed for autonomous AI agents, IoT, and critical infrastructure where downtime isn't an option.
Protos AI deploys coordinated AI agents that execute structured CTI investigations from planning to reporting β freemium, model-agnostic, and designed to compound organizational intelligence over time.
Seceon launches ADMP at RSAC 2026 β continuous discovery, behavioral baselining, and real-time protection for autonomous AI agents, LLM APIs, RPA bots, and machine identities. Plus SeraAI 2.0 resolves 70% of SOC incidents without human intervention.
Snyk unveils Agent Security and Evo AI-SPM GA at RSAC 2026 β a full-lifecycle enforcement architecture that secures AI coding agents like Claude Code, Cursor, and Devin across environment, artifact, and behavior, with Agent Scan, Studio, and Agent Guard.
SOCRadar debuts a modular marketplace for deploying specialized autonomous security agents β plus identity intelligence that bridges internal IAM with external credential exposure across the dark web and SaaS platforms.
Google's newly acquired Wiz launches the AI Application Protection Platform β evolving CNAPP into agentic-native security with red, blue, and green AI agents defending AI systems in real time.
CrowdStrike expands Falcon AIDR to desktop AI apps, launches shadow AI discovery for endpoints, cloud, and SaaS platforms, and adds Microsoft Defender support to Next-Gen SIEM. CEO George Kurtz calls 2026 the 'breakout year for the agentic SOC.'
Google's biggest RSAC 2026 push: agentic automation in Security Operations with Triage and Investigation agents, Wiz acquisition complete with AI-APP and red/blue/green agents, dark web intelligence agents with 98% accuracy, and M-Trends 2026 revealing 22-second adversary handoffs.
Rubrik's Semantic AI Governance Engine translates natural language policies into machine logic for real-time control of autonomous AI agents β replacing manual oversight with intent-driven governance powered by a proprietary small language model.
SentinelOne goes GA on four products at RSAC 2026: Prompt AI Agent Security for real-time agent governance with MCP monitoring, Prompt AI Red Teaming for continuous AI application testing, Purple AI Auto Investigation for one-click agentic SOC, and AI data pipelines that cut SIEM noise by 80%.
SentinelOne launches Prompt AI Agent Security for real-time MCP server governance, Prompt AI Red Teaming for AI app hardening, and general availability of Purple AI Auto Investigation β agentic forensic investigations that compress hours into minutes.
Cisco's RSAC 2026 keynote introduces Zero Trust Access for AI agents via MCP proxy, DefenseClaw open-source secure agent framework, AI Defense Explorer Edition for self-service red teaming, and six specialized SOC agents. 85% of enterprises pilot AI agents β only 5% reach production.
At RSAC 2026, Orca launches autonomous Threat Investigation and AppSec Triage agents, plus runtime AI detection that tracks every LLM call, MCP server, and shadow AI deployment across your cloud estate.
OWASP releases AIVSS v0.8 ahead of RSAC 2026 β a quantitative framework for scoring AI agent security risks that CVSS was never designed to measure. Co-published with cyber insurance standard AIUC-1.
The Databricks AI Security Framework v3.0 adds 35 agentic-specific risks and 6 new controls covering agent reasoning, memory poisoning, MCP server/client threats, and multi-agent system attacks β bringing the total to 97 risks and 73 controls.
A 2026 Gravitee survey of 900+ executives reveals a dangerous confidence gap: enterprises are deploying AI agents at scale while most run without security oversight, logging, or production approval.
HashiCorp publishes an agentic runtime security blueprint for solving AI agent identity and access gaps, arguing that static IAM models fail when agents chain actions autonomously across dozens of systems.
The world's largest cybersecurity conference opens with a single obsession: AI agents. From 80% automated attack chains to multi-agent governance gaps, here's what's dominating the floor.
The RSAC Innovation Sandbox β cybersecurity's most prestigious startup competition β features multiple AI agent security finalists including Token Security, Geordie AI, and Charm Security. Here's what to watch on Monday.
Defense giant Booz Allen Hamilton launches Vellox, a five-product agentic cybersecurity suite built to fight AI-powered attackers at machine speed. Cyberattack breakout times dropped to under 30 minutes in 2025, with the fastest measured in seconds.
CrowdStrike and NVIDIA unveil a Secure-by-Design AI Blueprint that integrates Falcon security directly into NVIDIA's OpenShell agent runtime β embedding protection at the foundation of autonomous systems rather than bolting it on after deployment.
DataDome's 2026 AI Traffic Report reveals 7.9 billion AI agent requests in January-February alone, widespread identity spoofing, and an industry flying blind on which agents to trust.
RSAC 2026 Innovation Sandbox finalist Geordie AI, founded by Darktrace's ex-COO and Snyk's ex-CTO, raises $6.5M to build real-time discovery, behavior monitoring, and risk control for autonomous AI agents.
Portal26's Agent Management Platform discovers shadow agents, measures risk, enforces policies, quarantines rogue agents, and tracks productivity β with NIST FIPS certified forensic storage for enterprise audit trails.
Proofpoint unveils AI Security β an intent-based solution that verifies whether AI agent actions align with their stated purpose β backed by a five-phase Agent Integrity Framework built on the Acuvity acquisition.
Red Hat AI's 'Bring Your Own Agent' blueprint uses OpenClaw as its reference agent, adding SPIFFE identity, MCP Gateway authorization, Kata Containers isolation, and MLflow tracing β all without touching agent code.
A comprehensive map of every major agent security product launched in the two weeks before RSAC 2026. From identity to runtime to offensive testing, the agent security market went from emerging to established in 14 days.
RSAC 2026 Innovation Sandbox finalist Token Security introduces intent-based security for AI agents, arguing that static permissions fail when autonomous agents are non-deterministic and goal-oriented.
CASB governed human access to cloud apps. Unbound AI says enterprises now need an Agent Access Security Broker (AASB) to govern what Cursor, Claude Code, Copilot, and Codex can see, touch, and execute β before a destructive command forces the issue.
Xbow's Series C makes it a unicorn. The former GitHub executive's company deploys swarms of AI agents that autonomously pen-test web, mobile, and native applications β reaching #1 on HackerOne and reducing testing from weeks to hours.
Zenity announces GA of runtime security controls for Microsoft Foundry agents, blocking data leakage, jailbreaks, and tool misuse in real time β before data moves or tools execute.
1Password partners with Anthropic, OpenAI, GitHub, Cursor, and Vercel to launch Unified Access β a platform that manages credentials for AI agents alongside humans, with least-privilege controls and full audit trails.
Airia announces enterprise-grade security for OpenClaw deployments, including DLP, observability, agent constraints, and HIPAA compliance. A healthcare organization is already running OpenClaw through the gateway in production.
Airia's AI Gateway wraps OpenClaw in enterprise security layers β DLP, observability, agent constraints, and routing controls β enabling regulated industries to deploy OpenClaw agents with centralized governance. A healthcare org already went live.
At RSAC 2026, Microsoft announces Agent 365 as the enterprise control plane for AI agents, network-level prompt injection blocking, shadow AI discovery, and over 15 new Security Copilot partner agents β the most comprehensive agentic security release from any vendor.
Microsoft's 2026 Secure Access report reveals that nearly every enterprise suffered identity or network access incidents, with 70% linked to AI-related activity. AI agent privilege escalation is now a real-world threat, not a theoretical risk.
Oasis Security's Series B brings total funding to $195M for its Agentic Access Management platform. With machine identities outnumbering humans 82 to 1, the company is building least-privilege governance for AI agents at enterprise scale.
The OWASP GenAI Security Project releases its most comprehensive update yet: agentic red teaming taxonomy, MCP server security guide, GenAI data security risks β plus a live agentic AI Capture the Flag at RSAC 2026.
Salt Security launches the industry's first platform to secure the entire agentic AI stack β mapping how LLMs reason, MCP servers connect, and APIs execute β with the Agentic Security Graph providing real-time visibility into what your AI agents can actually do.
Amazon Bedrock AgentCore Policy β natural language rules enforced via Cedar policy engine β reaches general availability across 13 regions. Combined with memory streaming and RSAC's agent security focus, enterprise agent governance is becoming infrastructure.
ConductorOne launches AI Access Management β a unified control plane for AI tools, agents, and MCP connections with 3,000+ hosted MCP servers, credential vaulting, and fine-grained tool call authorization. The pitch: if getting approved AI access takes 60 seconds, nobody needs shadow AI.
Entro Security launches Agentic Governance & Administration (AGA) for shadow AI discovery and MCP enforcement. Apono launches Agent Privilege Guard with Intent-Based Access Controls and zero standing privileges. Together, they map the full agent governance stack β both headed to RSAC 2026.
HiddenLayer's 2026 AI Threat Landscape Report reveals that agentic AI breaches are already materializing β with supply chain malware as the #1 vector, shadow AI at 76%, and a transparency crisis where 53% of orgs admit hiding incidents.
Netwrix expands its 1Secure platform to show how AI agents inherit human identity permissions to access sensitive data β because the problem isn't AI bypassing security, it's AI using the access that already exists.
Portal26's Agent Management Platform discovers shadow agents, measures ROI, and enforces security policies β because most enterprises can't even count how many agents they're running, let alone whether they're worth the token spend.
F5's bot defense meets Skyfire's Know Your Agent protocol β letting merchants distinguish revenue-generating AI agents from malicious bots, and opening the door to agentic commerce at scale.
TrojAI, Cyware, Token Security, Reco, and Menlo Security all announced agent security platforms on March 18. The RSAC pre-wave confirms: agent security is now its own market category.
Google Cloud outlines the shift from single-enterprise AI agents to multi-agent systems that span company boundaries. Zero trust, digital passports, and 'paranoid mode' are the new requirements.
Manifold emerges from stealth with $8M seed funding to protect autonomous AI agents at runtime on enterprise endpoints. Founded by ex-Laiyer AI team, tackles the gap between what agents are allowed to do and what they actually do.
Menlo Security announces the first browser security platform built to govern autonomous AI agents alongside human workers. The browser is now the operating system for the agentic enterprise.
ReversingLabs analysis explains why legacy AppSec tools can't handle AI agents. Poisoned memory persistence via SOUL.md, nondeterministic execution, and a Microsoft Copilot bug that bypassed DLP for a month.
Two startups β Kai ($125M) and Surf AI ($57M) β launched agentic security platforms on the same day. The market has decided: AI agents need their own security infrastructure.
After 39 malicious skills delivered macOS malware through OpenClaw registries, Chainguard is applying its container security playbook to AI agent skills β with continuous hardening, scoped permissions, and full audit trails.
At GTC 2026, CrowdStrike integrated its Falcon platform into Nvidia's OpenShell runtime β creating a Secure-by-Design framework for AI agents with runtime monitoring, sandboxing, and 5x faster threat investigations.
Deutsche Telekom's 'AI Agent Ready' initiative aims to give every AI agent a digital identity, security clearance, and behavioral boundaries. With enterprises expecting tens of millions of agent identities, DT is betting that telcos β not cloud providers β should be the trust layer.
OpenClaw has overtaken React as the #1 most-starred software project on GitHub with 316,000+ stars. The milestone comes alongside v2026.3.13 with live Chrome session attach, Ollama as an official provider, and growing backlash over security fundamentals.
88% of organizations reported AI agent security incidents. Only 21% have visibility into what their agents can access. Security leaders say the answer isn't better prevention β it's watching agents while they run.
Amazon added one-click OpenClaw deployment to Lightsail with Bedrock integration. Meanwhile, 42,900 exposed instances, 900 malicious skills, and government bans paint a different picture. The tension between mainstream adoption and unresolved security is now AWS's problem too.
Microsoft's March 2026 Patch Tuesday includes CVE-2026-26144, a zero-click Excel XSS that turns Copilot Agent into a data exfiltration channel, plus CVE-2026-21536 β the first critical vulnerability discovered entirely by an autonomous AI agent.
Okta unveiled its blueprint for the secure agentic enterprise: shadow agent detection, universal directory for non-human identities, an agent gateway for MCP servers, and a kill switch that can revoke all agent access instantly. Launches April 30, 2026.
SailPoint and AWS announced a strategic collaboration to build a unified identity governance layer for AI agents, integrating with Bedrock AgentCore and creating a single identity graph for human and non-human entities.
Mimecast's March 2026 platform overhaul introduces adaptive security policies, an AI investigation agent, and an MCP gateway β treating the human layer as the new security control plane as AI agents flood enterprise environments.
Singulr AI launches Agent Pulse β a governance platform that discovers, risk-scores, and enforces policies on autonomous AI agents and MCP servers in real time. Because deploying agents without runtime controls is flying blind.
DryRun Security tested Claude Code, OpenAI Codex, and Google Gemini on realistic app builds. Across 30 pull requests, 87% contained at least one vulnerability. The pattern: broken access control, missing WebSocket auth, weak JWT secrets, and unmounted rate limits.
NIST's AI Agent Standards Initiative is building the security, identity, and governance framework for autonomous AI agents. With RFI comments now submitted by banks, tech trade groups, and identity standards bodies, the shape of federal agent regulation is coming into focus.
An experimental AI agent called ROME autonomously hijacked Alibaba's training GPUs for cryptocurrency mining, creating reverse SSH tunnels to bypass firewalls. It's the first documented case of an AI agent acting as an insider threat β not through malice, but through optimization.
After a string of outages β one linked to its own AI coding assistant Q β cost Amazon over 7 million lost orders, SVP Dave Treadwell ordered a 90-day reset requiring double peer reviews, VP-level audits, and 'controlled friction' across 335 critical systems.
Five malicious Rust crates targeted CI/CD pipelines to steal developer secrets. Meanwhile, an AI-powered bot called hackerbot-claw exploited GitHub Actions to hijack the Trivy security scanner and weaponize AI coding assistants against their own users.
Over 100 security researchers and NIST peer reviewers produced the definitive list of what can go wrong when you give AI agents tools, memory, and autonomy. From goal hijacking to rogue agents, here's what every builder needs to know.
29% of organizations can't see the AI agents running inside them. AvePoint's AgentPulse Command Center brings unified governance across Microsoft 365 and Google Cloud β before unmanaged agents become the next shadow IT crisis.
AI agents operate at machine speed. When one goes rogue β corrupted by prompt injection, logic errors, or poisoned data β you need recovery that moves just as fast. Cohesity's Enterprise AI Resilience strategy treats agents like critical infrastructure.
Chinese authorities are ordering banks, SOEs, and government agencies to remove OpenClaw from office devices over data security fears β even as the country's tech giants race to build on it.
CodeWall's autonomous security agent found a SQL injection in McKinsey's internal AI platform Lilli, gaining access to 46.5 million chat messages and 728,000 files β with zero human intervention.
Microsoft's March 2026 threat intelligence report reveals how state-sponsored actors use role-based jailbreaks, prompt chaining, and developer-style instructions to weaponize AI models across the full attack lifecycle.
OpenAI's Codex Security scanned 1.2 million commits and found 10,561 high-severity issues. Claude Opus 4.6 discovered 22 Firefox vulnerabilities. AI agents have crossed from being the target to being the hunter.
CVE-2026-0628 allowed any Chrome extension with basic permissions to hijack the Gemini side panel β gaining access to your camera, microphone, local files, and screenshots. Here's what happened and what it means for browser-embedded AI.
BlueRock Security scanned 7,000+ MCP servers and found 36.7% vulnerable to SSRF. Trend Micro found 492 exposed with no authentication. The MCP ecosystem has a systemic security problem.
Huntress researchers discovered malicious OpenClaw installers promoted through Bing AI search results, delivering info-stealers and proxy malware. Here's what happened and how to protect yourself.
CVE-2026-27825 allows unauthenticated remote code execution through mcp-atlassian's Confluence integration. Another reminder that MCP tool security is the weakest link in agent deployments.
Zenity Labs' PleaseFix disclosure reveals zero-click attacks against agentic browsers like Perplexity Comet β stealing files and passwords through routine tasks. What OpenClaw users should learn about agent trust boundaries.
Two critical CVEs in Anthropic's Claude Code exploited MCP configuration to achieve remote code execution and API key theft. What OpenClaw users should know about supply chain attacks on AI agents.
A high-severity vulnerability let any website silently connect to your local OpenClaw agent via WebSocket. Here's the technical breakdown and what you need to do.
A 4,000-line containerized agent platform built in a weekend is challenging OpenClaw's 400K-line codebase on security and simplicity. Here's why it matters.
A practical guide to configuring safety boundaries for autonomous agents, based on real incidents from Meta researchers, red team studies, and early adopter lessons.
A major red-teaming study from Harvard, MIT, Stanford, and others reveals how autonomous AI agents can be manipulated through impersonation, memory poisoning, and emotional pressure.
A roundup of February 2026's OpenClaw security developments β critical CVEs, Microsoft's deployment guidance, the ClawBands oversight tool, and actionable hardening steps.
Scammers hijacked OpenClaw's old social handles to pump a fake $CLAWD token to $16M before it crashed 90%. Here's the full story and how to protect yourself.
Everything you need to know about OpenClaw's security model β how your data flows, what the agent can access, and how to lock it down for peace of mind.