Part 1 of a Two-Part Series on Third-Party Cyber Risk Management in the Age of Frontier AI
By: Chris Gordon/Senior Solutions Consultant
For years, organizations have invested heavily in strengthening their own cybersecurity environments. They have hardened infrastructure, improved identity controls, expanded vulnerability management programs, and introduced increasingly sophisticated detection and response capabilities.
But the next significant cyber risk may not originate inside the environment you control.
It may come through a vendor, software provider, open-source dependency, or AI-enabled service that your business relies on every day.
That isn’t a new cybersecurity problem. Third-party risk has existed for decades. What is changing is the speed, scale, and sophistication with which AI can identify and exploit weaknesses across an interconnected ecosystem.
And that changes the conversation for CISOs, CIOs, and business leaders.
The most advanced AI models are becoming increasingly capable of reasoning through complex problems, identifying vulnerabilities, chaining multiple weaknesses together, and executing tasks with less human intervention.
And this isn’t simply theoretical.
In Anthropic’s Project Glasswing, approximately 50 participating organizations used Claude Mythos Preview to identify more than 10,000 high- or critical-severity vulnerabilities in their software. Anthropic reported that Cloudflare alone identified 2,000 bugs across critical-path systems, including 400 rated high or critical.
Independent testing reinforces just how quickly these capabilities are advancing. The UK’s AI Security Institute found that Claude Mythos Preview successfully completed 73% of its expert-level cybersecurity capture-the-flag challenges. More significantly, it became the first AI model to complete the Institute’s 32-step simulated corporate-network attack from beginning to end.
These capabilities have obvious defensive value. Security teams can potentially identify weaknesses faster and at a scale that was previously impossible.
But the same technological progression fundamentally changes the economics of cybersecurity.
Finding vulnerabilities is becoming easier.
Determining which ones actually matter, prioritizing them, and fixing them fast enough is becoming the harder problem.
For defenders, AI creates enormous leverage.
For attackers, it can do the same.
And because no enterprise operates in isolation, that acceleration doesn’t stop at the corporate perimeter.
As large enterprises improve their own security posture, attackers have an obvious alternative: target the organizations they trust.
Mid-sized vendors, SaaS providers, outsourced service providers, and open-source projects often sit at critical points in the enterprise supply chain. Yet they may not have the people, budgets, processes, or advanced security capabilities available to their largest customers.
That creates a visibility gap.
An enterprise may understand its own vulnerabilities extremely well while having limited insight into the changing risk posture of the organizations connected to it. And AI potentially makes that gap more consequential.
Instead of asking only: “How secure is our environment?” security leaders increasingly need to ask: “Where are we exposed through the environments we don’t control?”
The economics are straightforward.
Why attack hundreds of well-defended enterprises individually if compromising one widely used provider, software dependency, or service can provide a path to many of them?
SolarWinds demonstrated the potential impact of this model years before today’s frontier AI capabilities emerged. More recent attacks involving SaaS platforms, stolen credentials, third-party integrations, and software dependencies continue to demonstrate the same fundamental principle.
The supply chain can provide attackers with scale. AI can potentially provide them with speed.
Together, those forces make third-party cyber risk a business resilience issue—not simply another security-control problem.
There is another dimension to the challenge.
AI isn’t only changing how vulnerabilities can be discovered and exploited. AI itself is becoming part of the third-party ecosystem.
Vendors are introducing AI capabilities into their products. Employees are connecting third-party AI tools to enterprise systems. Organizations are experimenting with autonomous agents, external models, APIs, SaaS applications, and new integrations at a rapid pace.
That creates a deceptively simple governance problem:
You can’t govern what you can’t see.
Security teams need to understand not only which third parties have access to their environments, but increasingly:
Traditional vendor inventories were not designed to answer those questions.
For many organizations, third-party cyber risk management still relies heavily on periodic questionnaires, attestations, certifications, and scheduled reassessments.
Those tools still have a role.
But a questionnaire represents a point in time.
Cyber risk does not.
A vendor that appeared secure during an assessment six months ago may be exposed to a newly discovered vulnerability today. A new AI integration may change its attack surface tomorrow. A fourth-party dependency may introduce exposure neither organization recognized when the original assessment was completed.
The gap between assessment and actual risk is becoming increasingly important.
Modern TPCRM programs therefore need to complement traditional assessment methods with more continuous intelligence and a clearer understanding of which exposures actually create material business risk.
This may be one of the most important changes AI brings to cybersecurity.
If AI dramatically increases the number of vulnerabilities organizations can discover, attempting to remediate everything becomes less realistic—not more.
Security teams need context.
The organizations best positioned for this environment will not necessarily be the ones that generate the most findings. They will be the ones that can see risk clearly, prioritize intelligently, and act quickly.
Cybersecurity is no longer contained behind the firewall, and neither is the CISO’s responsibility.
Third-party dependencies, SaaS platforms, cloud services, AI models, autonomous agents, APIs, and open-source software have created an environment in which enterprise risk extends far beyond infrastructure the organization directly owns.
That makes third-party cybersecurity a leadership and operating-model challenge as much as a technology challenge.
The goal isn’t to eliminate third-party risk. That isn’t realistic.
The goal is to understand where meaningful risk exists, make defensible decisions about it, and build the organizational capability to respond before an exposure becomes an incident.
In Part 2, we’ll look at what that operating model requires—and how organizations can move from periodic third-party assessments toward a more continuous, intelligence-led approach to cyber resilience.
Sources
