A sector-wide vendor audit lands with uncomfortable numbers
UpGuard's 2026 Higher Education Third-Party Cyber Risk report put a hard number on something security teams have suspected for years: the vendors universities depend on are themselves the weak point. Among the top 100 vendors that institutions most commonly use, 28 percent have had a data breach since 2024. Eleven percent currently show active infostealer infections, which the report identifies as a leading source of stolen credentials that later surface in bigger breaches.
UpGuard frames the core problem as a widening visibility gap, driven by complex vendor ecosystems and AI adoption that is moving faster than the review processes meant to govern it. That phrase matters because it reframes the question CIOs usually ask. The issue is not whether any single SIS, LMS or admissions tool is secure today. It is whether the institution, or the enterprise, actually knows the current state of every vendor it has already signed.
This year's incidents show the pattern, not just the headline
A rundown of 2026 campus breaches compiled by security firm Huntress makes the mechanism plain. At Newcastle University, a misconfigured connection tied to its admissions system exposed contact information for roughly 440,000 people, traced back to a single connection setting. At the University of Western Australia, an administrator left credentials for the Callista student information system publicly reachable. At Avans University of Applied Sciences in the Netherlands, a Power BI misconfiguration exposed personal data for nearly a year before anyone noticed.
The bigger incidents followed the same logic at scale. The ShinyHunters extortion group breached Instructure's Canvas LMS, used by roughly 9,000 schools, claiming 275 million records and later defacing Canvas login portals at hundreds of institutions during finals week. Roughly 90 days after that, the same group exploited an unpatched remote-code-execution flaw in Oracle PeopleSoft, reaching more than 300 instances at over 100 organizations, mostly universities. Huntress's summary of the pattern is blunt: it is rarely a sophisticated attack, and an open door and a stolen key tend to lead an attacker to the same place.
Decentralized buying is the root cause, not the AI layer
It would be easy to blame this year's wave on AI adoption outrunning security review, and UpGuard's report does cite AI exposure as a contributing factor. But the Huntress incident list tells a more structural story. Universities, like many large enterprises, let individual departments select and manage their own vendors and systems. Admissions offices buy their own CRM. Business schools buy their own analytics dashboard. Nobody owns the full inventory, so nobody notices when one piece of it opens up.
That decentralization is exactly the condition PE-backed SaaS and retail-tech enterprises recognize from their own org charts, where marketing owns its martech stack and HR owns its own learning platform outside IT's purview. The lesson from higher ed's breach year is that a point-in-time vendor review, done once at signing, is worthless against infrastructure that changes weekly. Supplier concentration and fragmented technology use are what let a single misconfigured connector become a 440,000-person exposure.
The identity layer is where the damage actually gets decided
Across nearly every incident in the Huntress rundown, the proximate cause was exposure, not exploitation: a setting left open, a credential left reachable, a dashboard left public. Even the Oracle PeopleSoft intrusion relied on an unpatched flaw rather than novel tradecraft. That consistency points CTOs toward where their defensive budget actually buys protection. Identity and access governance, not perimeter tooling, is where the real damage gets decided once an attacker is inside, regardless of whether the entry point was a misconfiguration or a stolen login, and regardless of how much was already spent on firewalls and endpoint tools that never came into play once that first door was left open.
For enterprise buyers of edtech and L&D platforms specifically, that means the procurement question has to extend past the vendor's own SOC 2 report into how that vendor's downstream integrations and admin accounts are governed. A vendor can pass every audit and still run a connector with a loose permission setting that nobody on either side is actively watching.
Continuous monitoring has to replace the annual review
UpGuard's recommendation, implicit in how it frames the visibility gap, is that institutions need continuous vendor monitoring rather than annual or point-in-time assessments. That is a real operational shift, not a compliance checkbox: it means tracking vendor security posture, credential exposure and configuration drift on an ongoing basis, the same way a security team watches its own network. Most university IT shops do not have the staffing for that today, and most enterprise L&D and procurement functions do not either, which is exactly why this gap keeps producing the same category of incident year after year regardless of how many new point solutions get added to the security budget.
The practical takeaway is that the vendor list itself needs to shrink before the monitoring problem becomes tractable. Every additional SaaS tool a department adopts outside central IT review adds another node that has to be continuously watched, and the Huntress and UpGuard data both suggest that is exactly where this year's breaches originated. Consolidation is a security control now, not just a cost-cutting exercise.
What this means for the enterprise AI and L&D stack
Enterprise buyers evaluating AI-enabled learning and training platforms should read this as a preview of their own exposure, not someone else's problem. Any AI tutoring tool, LMS plugin or analytics dashboard added to the stack this year inherits the same risk profile as the campus tools in UpGuard's audit: a vendor with its own sprawling sub-vendor network, adopted faster than security review can keep pace, often by a business unit that bypassed central procurement entirely.
The decision this forces onto the CTO's roadmap is concrete. Before approving the next AI-adjacent learning platform, confirm who owns continuous monitoring of that vendor's security posture after signature, not just at signature. If the answer is nobody, the institution or enterprise has already recreated the exact condition that produced this year's 440,000-record exposure and the 275-million-record Canvas claim. The fix is boring and organizational, not technical, which is precisely why it keeps getting skipped.
