AI Becoming an SOC Imperative for Curtailing Emerging Cyber Threats
At the DTX conference in Manchester, experts warned that AI is becoming a security operations centre (SOC) imperative for defending against autonomous cyber threats. While AI can automate log correlation, triage and reduce alert fatigue, panelists stressed that robust fundamentals—patching, access control, monitoring—must be in place first. Human oversight remains essential as AI models can hallucinate and attackers increasingly use AI for reconnaissance, phishing and malware. The shift also reshapes security roles, demanding skills like prompt engineering and GRC expertise.
Drupal Admins Rushing to Patch Maximum Severity SQL Injection Vulnerability
Drupal released an emergency patch on May 22 to close CVE‑2026‑9082, a maximum‑severity SQL injection flaw that impacts sites using PostgreSQL. The update also upgrades Symfony and Twig libraries, which are bundled in all supported Drupal branches (10.5, 10.6, 11.2,...
Why some Security Fixes Never Reach Your Vulnerability Dashboard
In late April 2026 a malicious Bitwarden CLI package was published to npm for about 90 minutes, stealing cloud and GitHub tokens from any developer who ran npm install. Bitwarden later issued CVE‑2026‑42994, but the CVE serves only as a retroactive alert...
Contractor’s Public GitHub Account Exposed GovCloud and CISA Credentials
An unsecured personal GitHub repository, labeled “Private‑CISA,” publicly exposed AWS GovCloud credentials, GitHub tokens, and internal CISA documentation. The repo, created by a contractor and live since November 2025, contained 844 MB of Kubernetes configs, scripts, and plain‑text passwords before being taken...
Internet Explorer May Be Dead, but Its Ghost Still Runs Malware
Microsoft’s legacy mshta.exe utility, a component of the retired Internet Explorer, remains a favorite living‑off‑the‑land binary for attackers. Bitdefender’s research shows MSHTA is being leveraged in active campaigns involving loaders like CountLoader, stealers such as LummaStealer, and the PurpleFox backdoor....
Microsoft May Security Patch Fails for some Due to Boot Partition Size Glitch
Microsoft’s May 2026 security update for Windows 11 can fail on devices with an EFI System Partition (ESP) of 10 MB or less, aborting at roughly 35‑36% during reboot. The failure leaves systems unpatched, exposing them to the dozens of security...
New Image-Based Prompt Injection Attack Targets Multimodal AI Models
Researchers at Xidian University unveiled CrossMPI, an image‑only prompt injection that subtly perturbs pictures to mislead multimodal AI models. The attack achieved a 66.36% success rate across five open‑source vision‑language models, outperforming prior methods by roughly 41 points and remaining...
AI Coding Is Fueling a Secrets-Sprawl Crisis Few CISOs Are Containing
AI‑assisted "vibe coding" is accelerating secret sprawl, as illustrated by Moltbook’s launch on Jan. 28, 2026, which exposed 1.5 million API tokens, 35,000 email addresses and private agent messages due to a misconfigured Supabase database. Researchers at Wiz and independent analyst...
Why the Best Security Investment a Board Can Make in 2026 Isn’t Another Tool
Boardrooms repeatedly approve new security tools, yet gaps persist because organizations lack true visibility into their environments. The article argues that the most valuable security capability in 2026 is a unified view of assets, access rights, and activity, not another...
Exchange Server Zero-Day Vulnerability Can Be Triggered by Opening a Malicious Email
Microsoft disclosed a zero‑day cross‑site scripting vulnerability (CVE‑2026‑42897) in on‑premises Exchange Server 2016, 2019 and Server Subscription Edition that can be exploited by simply opening a crafted email in Outlook Web Access. The flaw is already being used in the...
Autonomous Systems Are Finally Working. Security Is Next
Waymo recently surpassed 170 million autonomous miles without a serious crash, proving that speed, not perfection, drives success in complex systems. The article argues that security faces a parallel shift: detection capabilities have matured, but investigation latency now hampers defense. Lateral...
The Economics of Ransomware 3.0
Ransomware 3.0 has shifted from pure encryption to a triple‑extortion play that combines system lock‑out, data exfiltration, and pressure on customers, regulators and shareholders. The model, used by groups such as ALPHV and Cl0p, forces organizations to confront threats that...
Meet Fragnesia, the Third Linux Kernel Vulnerability in a Month
A new Linux kernel flaw named Fragnesia (CVE‑2026‑46300) has been disclosed, targeting the XFRM ESP‑in‑TCP subsystem to achieve local privilege escalation. The vulnerability provides an in‑memory write primitive that bypasses traditional file‑system permissions, allowing unprivileged users to corrupt security‑sensitive files...
FlowerStorm Phishing Gang Adopts Virtual-Machine Obfuscation to Evade Email Defenses
Security researchers have identified that the FlowerStorm phishing-as-a-service operation is now employing KrakVM, an open‑source JavaScript virtual machine, to obfuscate credential‑stealing code delivered via HTML attachments. The VM‑based payload encrypts malicious bytecode, evading static analysis and traditional email‑security tools while...
PraisonAI Vulnerability Gets Scanned Within 4 Hours of Disclosure
A critical authentication bypass in the open‑source AI orchestration framework PraisonAI was actively scanned less than four hours after a GitHub advisory disclosed it. The flaw, tracked as CVE‑2026‑44338, disables authentication by default in a legacy Flask API server and...
Fired Employee Sought AI Help to Hide Deletion of Hosting Firm’s Customer Data
A Virginia jury convicted former hosting‑firm employee Sohaib Akhter for using an AI chatbot to learn how to erase system logs and delete 96 federal databases after he and his brother were terminated. The brothers retained company laptops and privileged...
Google Discovers Weaponized Zero-Day Exploits Created with AI
Google's Threat Intelligence Group (GTIG) uncovered what it believes is the first AI‑crafted zero‑day exploit observed in the wild, a Python script that bypasses two‑factor authentication on a widely used open‑source system‑administration tool. The exploit was traced to a cybercrime...
AI Security Is Repeating Endpoint Security’s Biggest Mistake
AI security is repeating the endpoint security mistake of over‑relying on posture‑based controls. While organizations implement model inventories, SBOMs, and guardrails, they neglect behavioral detection that monitors actual AI actions. The article argues that, as with the shift from signature‑based...
8 Guiding Principles for Reskilling the SOC for Agentic AI
Top security leaders at DXC Technology, Accenture and former Virgin Atlantic CISO are pioneering the reskilling of SOC teams for agentic AI. They combine hands‑on sandbox environments, vendor‑led expertise and formal training tracks to embed AI agents into tier‑1 and...
Your CTEM Program Is Probably Ignoring MCP. Here’s How to Fix It
The piece highlights Model Context Protocol (MCP) as a hidden vulnerability in modern AI toolchains, warning that many security programs overlook it. It details real‑world breaches—such as a malicious npm MCP package that infected 300 firms and high‑severity CVEs that...
Pen Tests Show AI Security Flaws Far More Severe than Legacy Software Bugs
Penetration testing of AI and large language model (LLM) applications reveals a disproportionate share of high‑risk flaws. Cobalt’s 2026 State of Pentesting Report finds 32% of AI/LLM findings are high risk—about 2.5 times the 13% rate in traditional enterprise software—yet...
Your Refresh Plan Has a CVE Blind Spot
A healthcare client bought servers in 2017 and, due to COVID‑driven supply‑chain delays, extended the vendor’s end‑of‑life timeline to 2026 for software updates and 2028 for security patches. With new hardware now unavailable for up to a year and costs...
13 New Critical Holes in JavaScript Sandbox Allow Execution of Arbitrary Code
Thirteen critical vulnerabilities have been disclosed in the popular vm2 JavaScript sandbox, allowing attackers to break out of the container and execute arbitrary host commands. The most severe flaw, CVE‑2026‑26956, enables a full sandbox escape on Node 25 with WebAssembly support,...
Ollama Vulnerability Highlights Danger of AI Frameworks with Unrestricted Access
Researchers at Cyera uncovered a critical flaw (CVE-2026-7482) in the popular Ollama AI framework that enables unauthenticated attackers to trigger an out‑of‑bounds heap read via a crafted GGUF file. The bug can leak process memory—including prompts, API keys, environment variables,...
Poisoned Truth: The Quiet Security Threat Inside Enterprise AI
Enterprises are rapidly deploying internal LLMs, copilots, and autonomous agents, but security teams are overlooking a silent threat: AI data poisoning. The risk arises when models ingest corrupted, stale, or low‑quality data, causing them to produce plausible yet wrong outputs...
Train Like You Fight: Why Cyber Operations Teams Need No-Notice Drills
Cybersecurity detection has improved, yet response readiness still lags behind. The article contends that scheduled tabletop exercises cannot mimic the physiological stress of real incidents and recommends no‑notice drills to build instinctive, resilient response capabilities. Citing medical, military and psychological...
CISA Pushes Critical Infrastructure Operators to Prepare to Work in Isolation
U.S. Cybersecurity and Infrastructure Security Agency (CISA) unveiled CI Fortify, a national program urging critical infrastructure operators to plan for and operate in isolation from the internet and third‑party services during severe cyber incidents. The initiative emphasizes controlled disconnection, rapid...
AI Finds 20-Year-Old Bugs in PostgreSQL and MariaDB
AI‑driven security tool Xint Code uncovered a high‑severity heap overflow in PostgreSQL’s pgcrypto extension and a buffer‑overflow in MariaDB’s JSON schema validation, both tracing back to code written over two decades ago. The PostgreSQL flaws (CVE‑2026‑2005 and CVE‑2026‑2006) received CVSS...
The Fake IT Worker Problem CISOs Can’t Ignore
Hiring fake IT workers has become a widespread insider‑risk threat, with thousands of synthetic identities infiltrating U.S. firms. Amazon alone blocked more than 1,800 North Korean attempts to secure IT roles, while SentinelOne has logged 360 fake personas and over...
How CISOs Should Utilize Data Security Posture Management to Inform Risk
Chief information security officers face a persistent dilemma: they know their data security posture must improve, yet budgets and headcount are limited. Data security posture management (DSPM) platforms can deliver visibility into where sensitive data resides, but full‑featured solutions often...
Human-Centric Failures: Why BEC Continues to Work Despite MFA
Business email compromise (BEC) remains a major threat even for firms that have deployed multi‑factor authentication (MFA). Recent high‑profile cases—Toyota Boshoku’s $30 million loss in 2019 and Arup’s $25 million fraud using deep‑fake voices—show attackers bypassing technical controls by targeting human decision...
Just 34% of Cyber Pros Plan to Stick with Their Current Employer
A new IANS and Artico Search survey of 500 cybersecurity professionals reveals only 34% intend to stay with their current employer, underscoring a looming talent retention crisis for CISOs. While salary remains a factor, flexible hybrid work models—especially one to...
Managing OT Risk at Scale: Why OT Cyber Decisions Are Leadership Decisions
The article argues that operational technology (OT) cyber risk is fundamentally a leadership and governance issue, not just a technical one. OT environments differ from IT with long asset lifecycles, limited patching, and fragmented ownership, making consistent decision‑making across sites...
‘Trivial’ Exploit Can Give Attackers Root Access to Linux Kernel
A critical Linux kernel flaw dubbed Copy Fail (CVE‑2026‑31431) enables a trivial local exploit that grants root access by writing four arbitrary bytes to any readable file. The vulnerability, discovered by South Korea’s Theori, works on all major distributions released...
Bank Regulator Sounds Warning over Cybersecurity Threat Posed by AI Models
Australia’s prudential regulator APRA has issued a formal warning that frontier AI models such as Anthropic’s Claude Mythos could give cyber‑attackers unprecedented speed and precision against banks. The regulator’s letter highlights that existing governance treats AI as a routine technology, ignoring...
Dismantle Implicit Trust in OT Networks, CISA Tells Critical Infrastructure Operators
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) issued a 28‑page guide urging critical‑infrastructure operators to apply zero‑trust principles to operational technology (OT) networks. The guidance, co‑authored with the Department of Energy, the FBI, the Department of State and NIST,...
Max-Severity RCE Flaw Found in Google Gemini CLI
Security researchers at Novee Security disclosed a max‑severity (CVSS 10.0) remote code execution flaw in Google Gemini CLI and its associated GitHub Action. The vulnerability stemmed from implicit workspace trust in headless CI/CD environments, allowing malicious configurations to execute arbitrary commands....
SAP Npm Package Attack Highlights Risks in Developer Tools and CI/CD Pipelines
Supply chain researchers have uncovered a coordinated attack on SAP‑related npm packages, dubbed “mini Shai‑Hulud.” Malicious versions of mbt and several @cap‑js modules were published on April 29, embedding pre‑install code that harvested developer credentials, GitHub and npm tokens, and cloud...
Stopping the Quiet Drift Toward Excessive Agency with Re-Permissioning
The article warns that AI agents have moved from answering questions to executing multi‑step actions, and that excessive permissions are creating a hidden security gap. Gartner forecasts $2.5 trillion in AI spending and 40 % of enterprise applications embedding task‑specific agents by...
ODNI to CISOs on Threat Assessments: You’re on Your Own
The Office of the Director of National Intelligence’s 2026 Annual Threat Assessment pivots from a global, forward‑looking outlook to a homeland‑centric, operational report. It drops dedicated sections on China, Russia, Iran and North Korea and omits the infrastructure‑campaign tracking that...
AWS Leans on Prior Ingenuity to Face Future AI and Quantum Threats
AWS marks its 20‑year anniversary while confronting AI‑driven attacks and the looming quantum‑computing risk. The Nitro hardware platform gives AWS a “zero‑human” infrastructure, enabling isolated bare‑metal instances and protecting encryption keys. Early adoption of symmetric encryption means most data at...
Critical Cursor Bug Could Turn Routine Git Into RCE
Security researchers at Novee Security discovered a critical vulnerability (CVE‑2026‑26268) in the Cursor IDE that enables remote code execution when its AI agent autonomously runs Git commands on a malicious repository. The exploit leverages standard Git hooks and bare repositories;...
Securing RAG Pipelines in Enterprise SaaS
Enterprise SaaS platforms are embedding AI agents that rely on Retrieval‑Augmented Generation (RAG) to pull proprietary data from wikis, CRMs, code repos, and other sensitive sources. Recent high‑profile failures—such as the zero‑click EchoLeak exploit, vector‑database exposures, indirect prompt‑injection via public...
Stopping AiTM Attacks: The Defenses that Actually Work After Authentication Succeeds
Adversary‑in‑the‑middle (AiTM) phishing bypasses MFA by stealing session tokens after a legitimate login. While passkeys and phishing‑resistant MFA reduce credential theft, the real target becomes the bearer token that remains trusted across devices. The article outlines three practical controls—binding sessions...
AI Is Reshaping DevSecOps to Bring Security Closer to the Code
Artificial intelligence is fundamentally reshaping DevSecOps by embedding security controls directly into code‑generation tools, expanding vulnerability detection with large‑language‑model scanners, and automating remediation suggestions. AI‑assisted coding assistants now enforce policies at the point of creation, while LLMs identify logic flaws...
The ‘Manager of Agents’: How AI Evolves the SOC Analyst Role
AI is reshaping the Security Operations Center by turning Tier‑1 analysts from data‑gatherers into orchestrators of autonomous agents. Instead of manually investigating each alert, AI agents now query systems, correlate signals and build evidence chains in real time. Analysts shift...
CISA Last in Line for Access to Anthropic Mythos
Anthropic’s Claude Mythos, a bug‑hunting AI model, is being rolled out through a tightly controlled initiative called Project Glasswing. While the NSA and the Department of Commerce have received access, the Cybersecurity and Infrastructure Security Agency (CISA) remains excluded. Bloomberg...
3 Practical Ways AI Threat Detection Improves Enterprise Cyber Resilience
Enterprises are overwhelmed by thousands of low‑value security alerts, prompting a shift to AI‑driven threat detection. By building behavioral baselines for users, devices, identities and cloud workloads, AI reduces noise and surfaces genuine threats faster. Cross‑telemetry correlation and automated triage...
Offer Customers Passkeys by Default, UK’s NCSC Tells Enterprises
The UK National Cyber Security Centre (NCSC) is urging enterprises to make passkeys the default authentication method for consumer-facing services. Passkeys, built on FIDO2 standards, offer phishing‑resistant, password‑less login by leveraging device‑bound cryptographic keys. The guidance highlights that while passkeys...
Microsoft Issues Out-of-Band Patch for Critical Security Flaw in Update to ASP.NET Core
Microsoft released an out‑of‑band update (10.0.7) to fix a critical CVSS 9.1 vulnerability (CVE‑2026‑40372) introduced in the ASP.NET Core 10.0.6 Data Protection library. The flaw miscalculates the HMAC validation tag, allowing forged authentication cookies, tokens and other protected payloads across...