May 2026 First-Half Edition
Coverage window: May 1-15, 2026
Executive Overview
The first half of May confirmed a hard pattern: attackers are moving through the systems organizations rely on most.
The strongest signals of this period did not come from one sector or one attack type. They came from learning platforms, pharmaceutical supply chains, security vendors, AI infrastructure, CI/CD plugins, software packages, official installer paths, firewall portals, SD-WAN controllers, mobile-device management systems, hosting panels, Linux estates, collaboration trust paths, browser-level AI deployment, and industrial environments.
The issue is no longer only whether a company has been breached. The sharper question is whether the systems organizations trust to learn, build, secure, host, authenticate, automate, connect, manufacture, and operate have been governed with the seriousness their role now demands.
The lead confirmed incident is Instructure / Canvas. Instructure confirmed a breach involving Canvas user data, including names, email addresses, student ID numbers, and messages between users. ShinyHunters claimed a much larger theft affecting nearly 9,000 institutions and hundreds of millions of users.
The confirmed incident was serious on its own, but the escalation made it more important: attackers later defaced school login portals, disrupting access during finals, and Instructure said it had reached an agreement with the hackers to prevent release of stolen data. Canvas is not a peripheral education tool. It is core learning infrastructure for many schools and universities.
When a platform like that is compromised, the impact moves from privacy exposure into institutional continuity and public trust.
The second major signal is ransomware pressure against healthcare-adjacent and manufacturing supply chains. West Pharmaceutical Services disclosed a disruptive ransomware incident in early May, while Foxconn confirmed a cyberattack affecting some North American factories as a ransomware group claimed theft of sensitive data. These incidents matter because both organizations sit in broader supply chains where operational disruption can ripple beyond the directly affected environment.
The third pattern is the continued targeting of the security and developer ecosystem. Trellix confirmed unauthorized access to part of its source-code repository. Checkmarx faced another supply-chain incident when a modified Jenkins AST plugin was published to the Jenkins Marketplace. PyTorch Lightning had malicious PyPI versions published after publisher-account compromise. Mini Shai-Hulud reappeared across npm and PyPI packages, affecting developer ecosystems including TanStack and Mistral AI.
These events should be read together. Attackers are not only targeting applications. They are targeting the systems that build, secure, scan, and ship applications.
The fourth pattern is the acceleration of AI-related security risk. Braintrust urged customers to rotate AI provider API keys after unauthorized access to one AWS account. Google reported what it assessed as the first known AI-assisted zero-day exploit, targeting a two-factor authentication bypass in an open-source web administration tool. A fake OpenAI-branded Hugging Face repository reached high visibility while pushing infostealer malware.
Chrome’s automatic Gemini Nano download raised a governance and consent question around AI components landing on endpoints without clear user decision.
This directly extends the SAR line from prior Signal Watchtower work: AI is no longer only a tool to adopt. It is becoming infrastructure, attack surface, governance challenge, and attacker accelerant at the same time.
Confirmed Disclosures and Company-Acknowledged Investigations
Instructure / Canvas – Confirmed
Instructure confirmed a cybersecurity incident involving Canvas user data, including names, email addresses, student ID numbers, and messages among users. The company said there was no evidence that passwords, birth dates, government identifiers, or financial information were involved. ShinyHunters claimed responsibility and alleged a much larger scale, including data from nearly 9,000 schools and roughly 275 million individuals. Those larger figures remain attacker claims unless independently confirmed.
The incident escalated when attackers defaced Canvas login pages for multiple schools, turning data theft into public pressure. Instructure later said it had reached an agreement with the hackers, while acknowledging that it could not guarantee the data had been fully erased.
Why this matters: Canvas is not just a classroom tool. It is an institutional operating layer. It holds communications, student identifiers, course workflows, staff relationships, and school dependency. A compromise of this kind affects privacy, continuity, institutional trust, and the confidence of students, staff, parents, and administrators.
West Pharmaceutical Services – Confirmed Ransomware Disruption
West Pharmaceutical Services disclosed a disruptive ransomware incident in early May. The importance of this incident is its operational context. West is a major pharmaceutical packaging and healthcare-supply-chain company, and disruption in this environment raises questions about resilience, supplier continuity, and the cybersecurity posture of healthcare-adjacent infrastructure.
Why this matters: this is not simply a corporate ransomware case. It touches the broader healthcare and pharmaceutical supply chain, where downtime and operational disruption can create consequences beyond ordinary business interruption.
Foxconn – Confirmed Cyberattack, Ransomware Claim Under Assessment
Foxconn confirmed that some North American factories were affected by a cyberattack and that operations were resuming. A ransomware group claimed theft of 8 TB of sensitive data, including project files and technical information linked to major technology customers. The operational impact was confirmed, while the full data-theft claim remains tied to the attacker’s assertion unless further substantiated.
Why this matters: Foxconn is a critical manufacturing node in the global electronics supply chain. A cyberattack against this kind of supplier can create concern far beyond the directly affected facilities, particularly when attackers claim access to customer project documentation, schematics, or sensitive manufacturing data.
Trellix – Confirmed Source-Code Repository Access
Trellix confirmed unauthorized access to a portion of its source-code repository. The company said its investigation found no evidence that source-code release or distribution processes were affected, and no evidence that the source code had been exploited.
Why this matters: Trellix is a major cybersecurity vendor. Even when release processes are not affected, unauthorized access to a security vendor’s source repository is strategically relevant. Security vendors are no longer external observers of the threat landscape. They are targets inside it.
Braintrust – Confirmed AI Infrastructure Incident
Braintrust, an AI evaluation and observability startup, confirmed unauthorized access to an AWS account containing customer API keys used to access cloud-based AI models and urged customers to rotate keys. The importance of this incident is not the company’s size. It is the category: AI evaluation and observability platforms are becoming repositories of sensitive provider keys, model-access credentials, and operational telemetry.
Why this matters: AI infrastructure is becoming a new class of credential warehouse. A breach in this layer can create cost exposure, data exposure, model-abuse risk, and downstream operational risk even when the affected company is not a household name.
Checkmarx Jenkins AST Plugin – Confirmed Supply-Chain Incident
Checkmarx said a modified version of its Jenkins AST plugin had been published to the Jenkins Marketplace and advised users to verify they were running a trusted version. Public reporting described the incident as another TeamPCP-linked supply-chain intrusion.
Why this matters: Jenkins plugins sit close to build systems, source repositories, CI/CD secrets, cloud credentials, and deployment workflows. A compromised security plugin is not merely a vendor issue. It is a pipeline integrity issue.
SailPoint – Confirmed GitHub Repository Access
SailPoint disclosed unauthorized access to a subset of GitHub repositories, detected on April 20 and reported in early May. The company said it found no evidence that customer data in production or staging environments was accessed, and no service interruption was identified.
Why this matters: SailPoint sits in identity governance. Unauthorized access involving an identity-governance vendor’s repositories reinforces a wider May pattern: the vendors that help organizations manage trust are themselves part of the live attack surface.
Cushman & Wakefield – Confirmed Limited Vishing Incident
Cushman & Wakefield confirmed a limited data-security incident stemming from vishing, after ShinyHunters and Qilin separately made claims involving the company. The company said the incident was limited in scope and that it was communicating with impacted clients. A proposed class-action lawsuit alleged exposure of sensitive client information, but broader attacker and lawsuit claims should be separated from confirmed facts.
Why this matters: vishing remains one of the most effective intrusion paths because it converts human trust into system access. The incident fits a broader pattern of attackers targeting helpdesk, employee, SaaS, and support workflows where identity verification is often weaker than the systems those workflows protect.
NVIDIA GeForce NOW Armenia Partner – Confirmed Third-Party Exposure
NVIDIA confirmed that a reported GeForce NOW-related breach affected a regional Alliance partner in Armenia, not NVIDIA-operated services. Reporting says exposed data may include names, email addresses, phone numbers, dates of birth, usernames, membership status, and related account information, while passwords were not compromised.
Why this matters: the brand risk is global, but the breach path was regional partner infrastructure. This is the modern third-party operating model problem in miniature: customers see the global brand, while the actual exposure may sit inside local partner systems, delegated operations, or regional infrastructure.
Skoda Online Shop – Confirmed
Skoda confirmed unauthorized access to its online shop after attackers exploited a vulnerability in the software used to run the portal. Reporting indicates that potentially affected data included names, postal and email addresses, phone numbers, order information, usernames, and hashed passwords, while payment information was not compromised.
Why this matters: e-commerce portals remain rich sources of identity and transaction data. Even when financial data is separate, customer account and order data can support targeted phishing, impersonation, and fraud.
JDownloader – Confirmed Website and Installer Supply-Chain Incident
JDownloader’s official website was compromised between May 6 and May 7, causing certain Windows and Linux installer links to redirect users to malicious files. AppWork said the attacker changed CMS-managed web content and download links, but did not gain access to the underlying server stack. Other download channels, including macOS downloads, in-app updates, Flatpak, Winget, Snap packages, and the core JAR package, were not affected.
Why this matters: official download pages are high-trust distribution paths. Users do not have to visit a malicious website to be compromised if an official installer path is redirected. The incident reinforces the importance of code-signing checks, download provenance, and release-channel validation.
Critical Infrastructure and OT Signals
Polish Water and Infrastructure Exposure
Poland’s cyber environment continues to show why critical infrastructure and public-service systems remain a high-priority defensive concern. Earlier reporting described Russian-backed cyber activity targeting hospitals and water infrastructure, while Polish authorities have warned more broadly of growing sabotage and cyber pressure against national infrastructure.
Why this matters: water, energy, and public-service infrastructure are not abstract national-security categories. They are operational systems. When weak authentication, internet exposure, poor segmentation, or geopolitical targeting intersect, the risk can move from data theft to physical-world disruption.
Ransomware Context: Q1 Shows Consolidation, Not Relief
The Q1 ransomware picture does not support a slowdown narrative. Public reporting and threat-intelligence summaries point to a ransomware economy that remains highly active, consolidated, and increasingly oriented around access, exfiltration, and pressure rather than encryption alone.
Why this matters: West Pharmaceutical Services, Cushman & Wakefield, Canvas, and Foxconn all sit inside a broader extortion economy where stolen data, operational pressure, public defacement, leak-site pressure, and brand leverage can matter as much as encryption.
Hard signal: ransomware did not fade into fewer headlines. It matured into a concentrated, data-driven pressure system.
Software Supply Chain, Developer Ecosystem, and AI Package Risk
Mini Shai-Hulud – npm and PyPI Supply-Chain Campaign
Mini Shai-Hulud re-emerged in May across npm and PyPI. Reporting tied the campaign to affected packages in major developer ecosystems, including TanStack and Mistral AI. The campaign is important because it targets developer and CI/CD trust paths rather than only end-user systems.
Why this matters: modern software supply-chain attacks are not only about infecting end users. They are about stealing CI/CD secrets, compromising package-maintainer accounts, republishing malicious versions, and turning developer trust into propagation.
PyTorch Lightning – Malicious PyPI Package Versions
Malicious versions of the PyTorch Lightning package were uploaded to PyPI after compromise of the publisher account. Reporting says the malicious versions were designed to steal developer credentials and republish infected package versions using stolen tokens.
Why this matters: AI and machine-learning package ecosystems are now high-value credential-theft paths. Developer machines, cloud tokens, model-development workflows, and CI/CD environments all sit close to these packages.
Fake OpenAI Hugging Face Repository – AI Brand Trust Abuse
A malicious Hugging Face repository impersonated OpenAI’s legitimate Privacy Filter release, copied its presentation, and shipped malware. The repository appeared highly visible before removal, showing how AI-brand trust and model-sharing platforms can become malware delivery surfaces.
Why this matters: attackers are using AI-brand credibility, model-sharing platforms, and developer curiosity as delivery mechanisms. A model repository is no longer automatically safe because it looks popular, useful, or aligned with a trusted brand.
Exploitation and Vulnerability Roundup
Cisco Catalyst SD-WAN Controller CVE-2026-20182
CISA added Cisco Catalyst SD-WAN Controller CVE-2026-20182 to the Known Exploited Vulnerabilities catalog after confirmed exploitation in the wild. The vulnerability is reported as a CVSS 10.0 authentication-bypass issue affecting the SD-WAN control plane.
Why this matters: SD-WAN controllers are not ordinary network appliances. They sit in the control plane for distributed enterprise connectivity, influencing routing, segmentation, site-to-site communication, and network policy. An authentication-bypass flaw at this layer can give attackers access to the fabric that connects many systems, not merely one exposed server.
Hard signal: when the SD-WAN control plane is exposed, authentication bypass is not a device bug. It is a network-trust failure.
Palo Alto Networks PAN-OS CVE-2026-0300
Palo Alto Networks disclosed CVE-2026-0300, a critical buffer-overflow vulnerability in the PAN-OS User-ID Authentication Portal, also known as Captive Portal. Public reporting described limited exploitation and said the flaw could allow unauthenticated attackers to execute arbitrary code with root privileges on affected PA-Series and VM-Series firewalls when the portal is exposed to untrusted access.
Why this matters: firewalls are not ordinary perimeter devices. They are high-trust enforcement points. When exposed firewall services are exploitable, attackers can move from edge compromise to internal reconnaissance quickly.
Ivanti EPMM CVE-2026-6973
Ivanti patched CVE-2026-6973 in Endpoint Manager Mobile, an improper input validation vulnerability allowing a remotely authenticated administrative user to achieve remote code execution. Public advisories said Ivanti confirmed limited exploitation in the wild, and CISA added the issue to the Known Exploited Vulnerabilities catalog.
Why this matters: mobile-device management infrastructure is a high-leverage target. Compromise can affect devices, applications, identity, enterprise mobility controls, and administrative trust.
cPanel CVE-2026-41940
CVE-2026-41940 in cPanel and WHM was reported as actively exploited. Public reporting describes a critical authentication-bypass flaw involving CRLF injection behavior that can allow root-level compromise of hosting environments.
Why this matters: cPanel is widely used across web-hosting environments. A flaw at this layer can expose many downstream sites and hosted services through a single control-plane weakness.
Linux Privilege-Escalation Cluster: Copy Fail, Dirty Frag, and Fragnesia
The first half of May produced a serious Linux privilege-escalation cluster. Copy Fail, tracked as CVE-2026-31431, was added to CISA’s Known Exploited Vulnerabilities catalog after evidence of active exploitation. Dirty Frag was publicly disclosed with proof-of-concept exploit code and no patch available at the time of reporting. Fragnesia followed as another Linux page-cache corruption issue affecting the XFRM ESP-in-TCP subsystem.
Why this matters: Linux estates are not protected by reputation alone. Linux sits under cloud workloads, containers, CI/CD systems, web hosting, Kubernetes nodes, developer workstations, AI infrastructure, and security tooling. Local privilege escalation turns initial access into root access. When multiple such flaws surface in the same window, the defensive priority should shift from routine patch tracking to aggressive kernel update planning, workload isolation, and compromise review.
Hard signal: initial access is bad. Root access is strategic. May’s Linux flaw cluster is a reminder that containment fails when local escalation is treated as secondary risk.
Google AI-Assisted Zero-Day Exploit
Google’s Threat Intelligence Group reported what it assessed as the first known AI-assisted zero-day exploit. The exploit targeted a two-factor authentication bypass in an open-source, web-based administration tool. Google said it worked with the affected vendor and may have disrupted a planned mass exploitation campaign before launch.
Why this matters: this is a direct continuation of the SAR and Mythos line. The issue is not that every exploit will now be AI-generated. The issue is that AI-assisted vulnerability discovery is crossing from forecast into observed operational reality.
Hard signal: the danger is not that every bug is now AI-found. The danger is that AI is beginning to remove the discovery bottleneck.
PraisonAI – Late-Breaking AI-Agent Framework Watch
PraisonAI has surfaced in late vulnerability chatter as a possible AI-agent framework risk. At the time of drafting, no sufficiently authoritative public advisory, vendor statement, CVE record, or technical write-up has been confirmed in the reviewed material.
Why this matters: PraisonAI sits in the same broader risk category as MCP, Braintrust, fake OpenAI repositories, PyTorch Lightning, Mini Shai-Hulud, and other AI/developer ecosystem issues. Agentic AI frameworks often sit close to API keys, local execution, workflow automation, tool connectors, and developer environments. A serious vulnerability in that layer would not be just another package bug.
It would be another signal that AI orchestration frameworks are becoming part of the enterprise attack surface.
NGINX Rift / CVE-2026-42945 – Late-Breaking Watch
A late report claims NGINX Rift / CVE-2026-42945 is an 18-year-old heap-overflow vulnerability in NGINX with public exploit availability. At the time of drafting, this claim has not yet been corroborated through the broader set of official NGINX advisories, NVD/CVE indexing, CISA KEV listing, or major independent technical analysis available to this review.
Why this matters: if confirmed, this would be a major web-infrastructure risk. NGINX sits in reverse proxy, ingress, API gateway, load-balancing, web delivery, and cloud-native traffic paths across the internet. A confirmed exploited heap-overflow flaw in this layer would materially strengthen the May theme: trusted infrastructure layers are being tested from every direction.
Teams, ModeloRAT, and Collaboration Abuse
Reporting around hijacked Teams-themed delivery, ModeloRAT, trojanized Teams installers, ClickFix-style lures, and Teams impersonation chains points to the same pattern: collaboration and productivity brands are being used as trusted pathways into malicious execution.
Why this matters: workplace tools are trusted by default. That trust is now being weaponized. Attackers do not need every user to fall for suspicious malware links if they can move through familiar collaboration brands, helpdesk-style workflows, and productivity-tool prompts.
Chrome / Gemini Nano – AI Platform Governance Signal
Chrome’s reported automatic installation of the Gemini Nano on-device model has created a governance and consent issue. Reporting says Chrome can download a roughly 4 GB on-device AI model file to support features such as scam detection, writing assistance, and local AI APIs. Reports also noted that deleting the model file manually may not be durable because Chrome can download it again.
Why this matters: this is not a breach. It is an AI platform-governance signal. AI is entering endpoints through browsers, platform features, developer APIs, and default settings before many organizations have defined approval, inventory, risk ownership, or user-consent expectations.
Reader Action: How to Disable Chrome’s Local Gemini Nano Model
For non-technical users, the practical action is simple:
- Open Google Chrome.
- Go to Settings.
- Open System.
- Look for On-device AI.
- Turn On-device AI off.
- Restart Chrome and confirm the setting remains disabled.
Manual deletion of local files is not recommended for ordinary users. Reports indicate that Chrome may redownload the model if the underlying on-device AI setting remains enabled. For enterprise environments, this should be managed centrally through browser policy rather than by asking users to fix it one machine at a time.
The governance issue is not whether Gemini Nano is useful. The governance issue is whether AI components should appear on enterprise endpoints before users, administrators, and risk owners have made an explicit decision.
Signals and Patterns
Trusted Platforms, Broken Assumptions
The strongest May signal is that attackers continue to move through systems organizations assume are safe: learning-management platforms, security-vendor repositories, identity-governance vendors, AI evaluation platforms, Jenkins plugins, PyPI packages, npm packages, official software installers, firewalls, SD-WAN controllers, mobile-device management systems, browser AI components, and regional partner platforms.
The issue is not that all trust is misplaced. The issue is that trust must now be governed as a live security control, not inherited as an assumption.
Education Platforms Are Now Systemic Infrastructure
Canvas shows that digital learning platforms are no longer ordinary SaaS services. They hold communications, identity attributes, student identifiers, institution relationships, and operational workflows. When such platforms fail, the impact reaches students, staff, schools, universities, parents, and public education systems.
Education SaaS has become systemic infrastructure. It now needs systemic oversight.
Security Vendors Are Targets Inside the Threat Landscape
Trellix, SailPoint, and Checkmarx reinforce one message: security and identity vendors are not outside the threat model. Their repositories, build paths, plugins, and distribution channels are attractive because they sit close to customer environments and developer trust paths.
Defenders should not treat vendor security incidents as distant news. They should evaluate whether their own integrations, plugins, credentials, and downstream dependencies intersect with the affected vendor systems.
AI Infrastructure Is Becoming a Credential and Execution Layer
Braintrust, fake OpenAI repositories on Hugging Face, Google’s AI-assisted zero-day finding, PyTorch Lightning, Mini Shai-Hulud, and Chrome’s Gemini Nano controversy all point to one reality: AI is now part of the software supply chain, developer workflow, endpoint estate, browser layer, and attacker decision loop.
AI risk is no longer a future enterprise committee topic. It is already present in model repositories, browser settings, package registries, API keys, agent connectors, and open-source tooling.
Edge and Control-Plane Systems Remain Priority Targets
Cisco Catalyst SD-WAN Controller CVE-2026-20182, Palo Alto PAN-OS CVE-2026-0300, Ivanti EPMM, cPanel, and Linux privilege-escalation flaws show that edge and control-plane technologies remain high-value targets. Attackers understand that these systems offer leverage beyond one machine or one user account. They can open paths into identity, administration, hosting, remote access, routing, segmentation, and internal movement.
Patch availability is not risk reduction until exposure is reduced and compromise is ruled out.
Social Engineering Remains a First-Class Intrusion Path
Cushman & Wakefield’s vishing-linked incident and Teams-themed malware delivery both reinforce the same point. Attackers do not need to defeat every technical control if they can persuade a user, helpdesk operator, developer, or employee to complete the attack path for them. As I have often said, we humans are the greatest risk to cybersecurity.
Security training alone is not enough. Remote support flows, helpdesk verification, external collaboration, package publishing, and privileged workflow approvals all need stronger controls.
Ransomware Is Now a Pressure System
West Pharmaceutical Services, Canvas, Foxconn, and the Q1 ransomware data all point to a maturing ransomware economy. Encryption is no longer the only pressure point. Data theft, operational interruption, public defacement, leak-site pressure, customer anxiety, and supplier concentration all create leverage.
The ransomware story is no longer just “systems encrypted.” It is “trust converted into pressure.”
Defender Actions
Treat Trusted Dependencies as Live Attack Surface
Map SaaS platforms, AI tools, CI/CD plugins, browser AI features, official download paths, model repositories, package registries, identity-governance tools, learning platforms, regional partner systems, SD-WAN controllers, firewalls, MDM platforms, hosting panels, and security scanners as security-relevant dependencies. These should appear in architecture reviews, access reviews, incident playbooks, vendor-risk reviews, and monitoring scope.
Validate Build and Package Pipelines
Checkmarx, PyTorch Lightning, Mini Shai-Hulud, JDownloader, and the fake OpenAI Hugging Face repository all show that the route from code to package to runtime is now a prime target. Organizations should validate artifact provenance, package pinning, plugin integrity, signing, CI/CD secrets, package-registry accounts, and build-system permissions.
Reduce Control-Plane Exposure
Cisco SD-WAN, Palo Alto PAN-OS, Ivanti EPMM, cPanel, and Linux privilege-escalation flaws should push organizations to review externally exposed control-plane systems. Public access should be the exception, not the default. Administrative portals, captive portals, hosting panels, SD-WAN controllers, MDM consoles, and firewall management services require urgent exposure review.
Rotate and Scope AI Provider Credentials
Braintrust shows why AI provider API keys must be treated like privileged secrets. AI keys can carry cost, data, model access, and abuse potential. Store them narrowly, rotate them quickly, monitor usage spikes, and avoid placing broad provider credentials into tools without clear security controls.
Govern Browser and Endpoint AI Features
Chrome’s Gemini Nano controversy shows that AI can arrive through default browser behavior, not just formal enterprise adoption. Organizations should define whether local AI models, browser AI APIs, writing assistants, summarization features, and scam-detection models are approved, restricted, or centrally managed.
Reclassify Education SaaS as Critical Business Infrastructure
Canvas shows that learning platforms now require stronger vendor oversight, access control, incident-response planning, logging, contractual assurance, and data minimization. Schools and universities should treat LMS platforms as operational infrastructure, not merely academic software.
Separate Confirmed Incidents from Claims
Canvas, Cushman & Wakefield, Foxconn, PraisonAI, and NGINX Rift show why disciplined classification matters. Confirmed facts, attacker claims, lawsuit allegations, leak-site statements, and late technical reports must be tracked separately. That discipline protects credibility and prevents unverified narratives from becoming premature conclusions.
Closing Note
The first half of May did not produce one defining event. It produced a defining pattern.
The systems organizations rely on to learn, build, secure, authenticate, host, route, manage, browse, manufacture, and automate are now the systems attackers are testing first.
Canvas shows the risk of systemic education platforms. West Pharmaceutical Services and Foxconn show the operational stakes of ransomware in supply chains. Trellix, SailPoint, Checkmarx, PyTorch Lightning, Mini Shai-Hulud, and JDownloader show that developer and security ecosystems remain high-value targets. Cisco SD-WAN, Palo Alto PAN-OS, Ivanti EPMM, cPanel, and the Linux flaw cluster show that control-plane and infrastructure systems remain priority targets.
This is not a call for panic. It is a call for sharper trust management.
In the next phase of cyber risk, the most important question may not be whether a system is trusted. It may be whether that trust has been earned, scoped, monitored, and continuously revalidated.
Originally published on LinkedIn: www.linkedin.com/pulse/edition-007-2026-venkat-mangudi-cj8tc