AI Security
Cybersecurity innovations, AI-powered threat detection, responsible disclosure, and the defenders shaping the digital frontier.
81 articles
Cohesity and CISA Sign a Cybersecurity Information Sharing Partnership — AI-Driven Threat Detection Joins the Joint Cyber Defense Collaborative
Cohesity and CISA announced a voluntary cybersecurity information sharing partnership on May 21, 2026 — adding Cohesity's AI-powered data security telemetry to the Joint Cyber Defense Collaborative.
Cisco Refines Vulnerability Disclosure for the AI Era — Risk-Based PSIRT and AI-Accelerated Patching
Cisco published a new risk-based vulnerability disclosure approach on May 25, 2026 — using frontier AI models to find and fix flaws faster while concentrating advisories on the highest-risk issues.
NSA Publishes Its First MCP Security Playbook — How to Deploy Model Context Protocol Safely for AI Agents
The NSA's Artificial Intelligence Security Center released MCP security design considerations on May 20, 2026 — a 17-page Cybersecurity Information Sheet outlining how to safely deploy Model Context Protocol in agentic AI systems.
OpenAI Launches Daybreak — Frontier AI Models Get a Cybersecurity Initiative Focused on Vulnerability Detection and Patch Validation
OpenAI launched Daybreak in May 2026 — a cybersecurity initiative that pairs frontier AI models with Codex Security tooling to help organizations find vulnerabilities and validate patches before attackers strike.
Microsoft Agent 365 Adds Shadow AI Discovery and a Purview Claude Connector — Local Agents Finally Get Governance
Microsoft expanded Agent 365 on May 21, 2026 — new shadow AI discovery for local Windows agents, a Purview connector for Anthropic Claude, and Defender plus Intune controls bring previously invisible AI agents into enterprise governance.
Microsoft Open-Sources RAMPART and Clarity — A Safety-First Toolkit for AI Agent Development
Microsoft released RAMPART and Clarity as open-source tools on May 20, 2026 — operationalizing AI agent safety by turning red-team findings into repeatable tests and documenting design assumptions in agent workflows.
Anthropic Opens Glasswing — Mythos Cyber Findings Can Now Be Shared With the Wider Defensive Community
On May 19, 2026, Anthropic updated the Glasswing program so partners can now share Mythos-derived cyber findings, tools, and code with the wider defensive community for maximum security impact.
Microsoft's MDASH Multi-Model Agentic Security System Finds 16 Windows Flaws and Tops CyberGym at 88.45%
Microsoft unveiled MDASH on May 12, 2026 — a multi-model agentic security system built by the Autonomous Code Security team that found 16 new Windows vulnerabilities and scored 88.45% on the CyberGym benchmark.
OpenAI Launches Daybreak — Codex Security and Three GPT-5.5 Variants Team Up to Find and Patch Vulnerabilities
OpenAI launched Daybreak on May 12, 2026 — a cybersecurity initiative pairing Codex Security with three GPT-5.5 variants and eight major partners including Cisco, Cloudflare, CrowdStrike, and Palo Alto Networks.
Palo Alto Networks' May Patch Wednesday Goes AI-Native — 26 CVEs Found by Frontier Models in a Single Sweep
Palo Alto Networks disclosed 26 CVEs across 75 issues on May 14, 2026 — the first Patch Wednesday where the majority of findings came from frontier AI models scanning the company's own code.
Palo Alto Networks Unveils Idira — A Next-Gen Identity Security Platform Built for the Agentic AI Enterprise
Palo Alto Networks unveiled Idira on May 12, 2026 — a next-generation identity security platform that discovers, controls, and governs every human and AI agent identity across the modern enterprise.
Microsoft Unveils MDASH — A Multi-Model Agentic Scanning Harness That Tops the Vulnerability Benchmark
Microsoft announced MDASH on May 12, 2026 — a multi-model agentic scanning harness that delivers state-of-the-art results on a leading industry vulnerability discovery benchmark and orchestrates frontier models for defensive cybersecurity.
Cognizant Launches Secure AI Services — A Build-Time and Run-Time Trust Platform for Agentic Enterprise AI
Cognizant launched Secure AI Services on May 7, 2026 — a new integrated offering that combines a Secure Agent Development Lifecycle, Neuro Cybersecurity, and Responsible AI to govern and scale enterprise agentic systems.
OpenAI Launches Daybreak — A Codex Security Platform for AI-Powered Vulnerability Detection With Tier-1 Partners
OpenAI launched Daybreak on May 11, 2026 — a cybersecurity platform built on Codex Security and GPT-5.5-Cyber for vulnerability detection, threat modeling, and patch validation with Cisco, Cloudflare, CrowdStrike, and more.
VIAVI Launches CyberFlood CF1000 — Native 400G Security and AI Inference Fabric Validation at Multi-Terabit Scale
VIAVI Solutions launched the CyberFlood CF1000 on May 5, 2026 — a native 400G security and application performance test platform built to validate AI inference fabrics and multi-terabit data center infrastructures.
Palo Alto Networks Launches Frontier AI Defense — Autonomous Remediation Against Machine-Speed Attacks
Palo Alto Networks unveiled Frontier AI Defense on May 7, 2026 — a platform pairing AI-native security with Unit 42 expertise and a partner alliance to remediate machine-speed AI attacks in real time.
OpenAI Launches GPT-5.5-Cyber for Vetted Defenders — A Specialized AI Security Model
OpenAI's new GPT-5.5-Cyber, announced May 7, 2026, gives vetted cybersecurity defenders a specialized AI model for red teaming, penetration testing, and bug discovery via the Trusted Access for Cyber program.
Operant Launches Endpoint Protector to Defend Every AI Tool, Agent, and MCP Workflow
Operant AI's new Endpoint Protector, launched May 8, 2026, gives security teams a single endpoint surface for discovering and defending AI tools, coding agents, and Model Context Protocol workflows.
Intel 471 Launches Retroactive Threat Detection to Operationalize Intelligence in EDR and SIEM
Intel 471's new Retroactive Threat Detection (RTD) launched May 6, 2026 — turning new threat reports into instant queries against historical EDR and SIEM data to confirm or rule out compromise in minutes.
Synack's Sara Agentic AI Pentesting Hits General Availability for Continuous Security Validation
Synack announced general availability of Sara — its Synack Autonomous Red Agent — on May 5, 2026, combining agentic AI with human red-team validation for continuous penetration testing across the full attack surface.
Five Eyes Agencies Release First Joint Guidance for Securing Agentic AI Systems
On May 1, 2026, CISA, NSA, and four allied cybersecurity agencies published the first joint Five Eyes guidance on safely deploying agentic AI — a 30-page playbook organized around five risk categories for autonomous AI agents.
Microsoft Lays Out a Pre-Deployment Playbook for Frontier AI Security
Microsoft published a detailed pre-deployment AI security playbook on May 1, 2026 — Brad Smith and Natasha Crampton's blueprint for how frontier AI developers, governments, and deployers should secure the next generation of agentic models together.
Wiz × Lovable Brings Embedded Security Scanning Into the Vibe Coding Workflow
Wiz announced a new integration with Lovable that runs security scanning directly inside the vibe coding platform — vulnerabilities, secrets, and misconfigurations now surface in Lovable's built-in security view as part of the May 2026 GA rollout.
OpenAI Scales Trusted Access for Cyber With GPT-5.5-Cyber for Verified Defenders
OpenAI announced the expansion of its Trusted Access for Cyber program on April 30, 2026 with GPT-5.5-Cyber — a fine-tuned defender-focused frontier model now available to thousands of verified individual defenders and hundreds of teams.
Geordie AI Crowned RSAC 2026 Innovation Sandbox Winner for AI Agent Governance
Geordie AI took the Most Innovative Startup crown at the RSAC 2026 Innovation Sandbox in San Francisco — its AI agent governance and observability platform gives enterprises real-time visibility into their agentic footprint.
CISA Drops a Joint Zero Trust Playbook for Operational Technology — Aligned With NIST CSF 2.0
CISA, with partner agencies, published 'Adapting Zero Trust Principles to Operational Technology' on April 29, 2026 — a practical Zero Trust roadmap for OT operators aligned with NIST CSF 2.0 and ISA/IEC 62443.
Defenders Are Pulling Ahead in 2026 — Record Security Budgets and AI-Powered Network Intelligence
CyberEdge's 2026 Cyberthreat Defense Report shows 90% of organizations raised IT security budgets — and AI-powered network intelligence is tilting the defender-attacker balance.
Microsoft Patches an Entra ID Agent Identity Flaw Before AI Agents Could Be Hijacked
Microsoft shipped a patch on April 28, 2026 for a privilege escalation flaw in Entra ID's Agent ID Administrator role — closing a service principal takeover path that targeted AI agent identities.
Google Cloud Next '26 Unveils Threat Hunting Agents and an Agent Gateway for AI-Era Security
Google Cloud Next 2026 introduced a Threat Hunting agent, a Detection Engineering agent, and Agent Gateway for policy enforcement — extending Google's agentic defense stack for enterprises.
Microsoft Rolls Out AI-Powered Defense Stack and Joins Project Glasswing for Vulnerability Hunting
Microsoft's new Secure Now dashboard, AI-driven SDL upgrades, and Project Glasswing collaboration with Anthropic give defenders an AI-accelerated toolkit for the modern threat landscape.
Microsoft's Zero Day Quest 2026 Pays $2.3M to Researchers Who Hardened Cloud and AI
Microsoft's Zero Day Quest 2026 awarded $2.3 million across 80+ high-impact cloud and AI vulnerabilities — turning hacker creativity into a stronger Secure Future Initiative.
Vodafone and Google Cloud Bring Enterprise AI Security and Gemini Concierge to European SMBs
Vodafone and Google Cloud debut AI-powered managed security and Gemini Concierge for European SMBs — enterprise-grade protection launching first in Germany.
IBM Autonomous Security Uses Multi-Agent AI to Fight Back Against Agentic Cyberattacks
IBM Autonomous Security coordinates multi-agent AI across your full security stack to detect, analyze, and contain agentic cyber threats at machine speed.
92% of Enterprises Lack Visibility Into AI Identities — Here Is How to Fix It
Saviynt's 2026 CISO AI Risk Report finds 92% lack full AI identity visibility and 95% can't detect misuse. The research maps the governance gaps and a practical framework to close them.
Microsoft's Zero Trust for AI Framework: Securing Agentic Workloads in 2026
Microsoft's ZT4AI framework from RSAC 2026 gives enterprises a principled blueprint for securing AI agents, LLMs, and autonomous AI workloads using Zero Trust architecture.
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.
Google Gemini Blocked 8.3 Billion Harmful Ads in 2025 and Is Expanding in 2026
Google's annual ad safety report reveals Gemini AI blocked 8.3 billion harmful ads in 2025, catching 99%+ of violations before serving — and plans to expand to more ad formats this year.
OpenAI's GPT-5.4-Cyber Puts Advanced AI in the Hands of Verified Security Defenders
OpenAI launches GPT-5.4-Cyber for vetted security teams — a fine-tuned defensive model with binary reverse engineering capabilities and expanded access through the Trusted Access for Cyber program.
IBM Launches Autonomous Security Platform to Defend Against AI-Powered Cyber Attacks
IBM's new Autonomous Security platform uses multi-agent AI to detect and respond to agentic cyber threats at machine speed — a major defensive leap as attackers begin weaponizing frontier AI models.
Anthropic's Project Glasswing: $100M and 50 Partners to Patch Zero-Days With AI
Anthropic commits $100M and 50 industry partners to Project Glasswing — using Claude Mythos Preview to find and patch zero-day vulnerabilities across every major OS and browser.
Agentic AI Is Giving Security Operations Centers Their Edge Back
Enterprises running agentic AI in their SOC see 40% faster threat detection — as autonomous investigation and response tools compress human-driven timelines from hours to minutes.
AI Browser Extensions Are Your Enterprise's Biggest Unmanaged Security Surface
LayerX's 2026 Browser Extension Security Report reveals AI extensions carry 60% more CVEs and access cookies 3x more than standard extensions — with 15% of enterprise users already exposed.
Tanium Guardian Shines a Light on Shadow AI Across Every Endpoint
At RSAC 2026, Tanium unveiled Guardian Spotlight AI Tools — giving security teams real-time visibility into every AI model, agent, and MCP server running across endpoints.
Cynomi's CISO Intelligence Agents Give Every MSP a Full Virtual Security Team
Cynomi launched four AI co-worker agents on April 8 that act as a virtual CISO team — handling strategy, compliance, execution, and board reporting for MSPs.
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.
Project Glasswing: How Anthropic's AI Uncovered Decade-Old Vulnerabilities Across Every Major Platform
Anthropic's Project Glasswing used Claude AI to discover thousands of critical vulnerabilities — including a 27-year-old OpenBSD flaw and 16-year-old FFmpeg bug — across major OS and browser codebases.
Exabeam Now Monitors AI Agents Like It Monitors Human Users — With Behavioral Baselines
Exabeam's April 2026 New-Scale update introduces Agent Behavior Analytics, treating ChatGPT, Gemini, and Copilot integrations as first-class security identities requiring behavioral monitoring.
Microsoft Releases Zero Trust for AI: A Security Framework Built for the Age of AI Deployment
Microsoft's new Zero Trust for AI initiative delivers reference architectures, assessment tools, and governance workshops specifically designed to secure AI systems in enterprise environments.
The 1H 2026 AI Security Report: The Era of Agentic Security Has Arrived
The 1H 2026 AI and API Security Report finds enterprises racing to secure AI agents as agentic architectures reshape the enterprise attack surface.
ISC2 Weaves AI Security Into All Its Cybersecurity Certifications
ISC2 published AI security exam guidance on April 2, mapping AI concepts across 50+ domains in its entire certification portfolio for 265,000 members.
CSA's 2026 AI Cybersecurity Survey: 92% of Security Leaders Are Concerned About AI Agents
The Cloud Security Alliance surveyed 1,500+ security leaders for its 2026 report, finding near-universal concern about AI agent security alongside meaningful year-over-year progress in AI-powered defenses.
CrowdStrike, AWS, and NVIDIA Pick 35 AI Security Startups for Their 2026 Accelerator
Three of tech's heavyweights selected 35 AI-native security startups from hundreds of global applications for their prestigious 8-week cybersecurity accelerator announced March 24.
Cybersecurity Spending Tops $244 Billion in 2026 as AI Reshapes Defense
Gartner projects global information security spending will hit $244 billion in 2026, driven by AI-powered threats, cloud security at 28.8% growth, and post-quantum cryptography preparation.
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.
Google Cloud Launches an AI-Powered Agentic SOC With Mandiant Threat Intelligence
At RSAC 2026, Google Cloud unveiled an Agentic SOC where autonomous AI agents detect and respond to threats at machine speed, powered by Mandiant's frontline intelligence.
PwC's 2026 Threat Report: AI-Powered Identity Defense Is Outpacing the Attackers
PwC's Annual Threat Dynamics 2026 shows how security teams are deploying AI-powered identity defenses to counter automated attack tooling — and the defensive tools are keeping pace.
Microsoft Brings Zero Trust to AI Agents With Agent 365 and New Framework
Microsoft unveils Zero Trust for AI and Agent 365 at RSAC 2026, giving enterprises unified visibility, shadow AI detection, and governance over AI agents at scale.
Agentic AI Defense Dominates the 2026 Cybersecurity Excellence Awards at RSA
The 2026 Cybersecurity Excellence Awards at RSA crowned AI security as its most competitive category, recognizing purpose-built tools for governing and protecting autonomous AI agents.
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.
Sysdig Catches Langflow AI Platform Exploitation Within 20 Hours of Disclosure
Attackers reverse-engineered exploits for CVE-2026-33017 in Langflow within 20 hours of the advisory, with no public PoC available. Sysdig tracked 6 IPs across 48 hours.
An AI Agent Just Found Its First Critical CVE — XBOW Autonomously Discovers a 9.8-Severity Microsoft Vulnerability
XBOW, a fully autonomous AI penetration testing agent, independently discovered CVE-2026-21536, a critical RCE flaw in a Microsoft service — marking a milestone for AI-powered defense.
International Coalition Dismantles Four Record-Breaking IoT Botnets That Hijacked 3 Million Devices
The DOJ, Canadian, and German authorities take down Aisuru, Kimwolf, JackSkid, and Mossad botnets that powered 30+ Tbps DDoS attacks targeting hundreds of thousands of victims.
Perseus Android Malware Evolves From Cerberus to Steal Financial Data From Notes Apps
Security researchers discover Perseus, a Cerberus descendant that monitors note-taking apps for passwords and financial details, distributed through fake IPTV streaming apps.
GlassWorm Returns With a Second Wave — The Supply Chain Attack Expands From GitHub to npm Packages and VSCode Extensions
The invisible Unicode malware campaign that hit 151 Python repos has evolved, with security researchers detecting coordinated injections across npm, GitHub, and VSCode/OpenVSX extension marketplaces.
NVIDIA Open-Sources NemoClaw — A Security-First Stack for Deploying Autonomous AI Agents on Any Hardware
Built on the OpenClaw platform, NemoClaw bundles Nemotron models with sandboxed execution and privacy controls, enabling secure AI agent deployment from RTX laptops to DGX clusters.
GlassWorm Supply Chain Attack Hides Malware in Invisible Unicode Across 151+ GitHub Python Repos — Here's How Defenders Caught It
Security researchers at Aikido discover a novel attack using stolen tokens and invisible PUA Unicode characters to inject undetectable payloads into Django apps, ML code, and PyPI packages.
OpenAI's Codex Security Scanned 1.2 Million Commits and Found 10,561 High-Severity Vulnerabilities in Open-Source Projects
The AI-powered security agent discovers critical bugs in OpenSSH, Chromium, PHP, and GnuTLS during its research preview, with false positive rates dropping 50% over 30 days.
Europol Dismantles SocksEscort — A Global Botnet That Enslaved 369,000 Routers Across 163 Countries for Fraud
Operation Lightning takes down the SocksEscort proxy service, seizing 34 domains and freezing $3.5M in crypto after the botnet enabled millions in fraud through hijacked home routers.
Google Closes Its $32 Billion Wiz Acquisition — The Largest Deal in Google's History Reshapes Cloud Security
After a year of regulatory review, Google officially completes the all-cash acquisition of Wiz, which will join Google Cloud while maintaining multi-cloud security support.
Microsoft's March Patch Tuesday Fixes 2 Zero-Days and 79 Vulnerabilities — Including Critical Office Preview Pane Bugs
The March 2026 Patch Tuesday addresses 79 flaws across Windows, Office, and SQL Server, with two publicly disclosed zero-days and three Critical-rated remote code execution bugs.
Mandiant Founder Kevin Mandia Raises $190M for Armadin — An AI Security Startup That Hunts Threats Autonomously
After selling Mandiant to Google for $5.4B, Kevin Mandia is back with Armadin — an autonomous AI security agent platform backed by Accel, Google Ventures, and Kleiner Perkins.
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.
Amazon Introduces Mandatory Senior Review for AI-Generated Code Deployments After Learning Hard Lessons
After incidents involving autonomous AI coding tools, Amazon now requires senior manager sign-off before junior engineers can deploy AI-assisted code to production environments.
Claude AI Discovers 22 Security Vulnerabilities in Firefox in Just 14 Days — Ushering In AI-Powered Bug Hunting
Anthropic's Claude Opus 4.6 found 22 CVEs in Firefox during a two-week audit with Mozilla, including 14 high-severity flaws that traditional fuzz testing had missed.
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.
Pakistan-Linked APT36 Floods Indian Government Networks With AI-Generated 'Vibeware' Malware
Bitdefender reveals a nation-state campaign using AI coding assistants to mass-produce disposable malware in 12+ languages, overwhelming defenders with sheer volume.
Google's Threat Analysis Group Exposes 'Coruna' — An iOS Exploit Kit Packing 23 Zero-Day Vulnerabilities
Google TAG reveals a sophisticated commercial spyware vendor using a chain of 23 previously unknown iOS exploits to compromise iPhones without user interaction.
An AI-Powered Attacker Just Breached 600+ FortiGate Firewalls Using Automated Exploit Chains
Security researchers trace a mass exploitation campaign against Fortinet firewalls to an attacker using large language models to automate vulnerability discovery and exploit generation.
Push Security Ships Automatic Detection and Blocking for Malicious Browser Extensions
Push Security's new capability identifies and neutralizes known-malicious browser extensions before they can exfiltrate data or hijack sessions.
Federal Agencies Are Going All-In on AI for Cyber Defense — And the Results Are Already Showing
U.S. government agencies deploy AI-powered threat hunting and automated diagnostics as CISA's modernized defense strategy takes shape.