What Organizations Should Do Next
Managing AI supply chain risk takes more than a single cyber rating or a one-time vendor assessment.
The findings in this report reveal several distinct challenges: critical vulnerabilities across AI suppliers, signs of malicious activity, stolen corporate credentials, shared software dependencies, and integrations that can provide access to customer systems.
These risks require different responses. Some can be identified and monitored through external cybersecurity intelligence. Others require direct engagement with suppliers to understand the technologies they use, the information they process, and the access their products receive.
The following six actions connect Black Kite’s research findings to practical steps organizations can take to identify, prioritize, and reduce AI supply chain risk.
Layer 1: The AI company
01 Identify AI Suppliers and Understand Their Role
of AI companies have a limited publicly visible digital presence.
- Flag every supplier that provides AI capability, including AI features built into products not marketed as AI services.
- Record what each one does: what the product does, what information it processes, which business operations depend on it, and whether it reaches into your systems.
- Sort suppliers by role before you assess them. A model provider, a company building custom AI for you, and a packaged AI application carry different risks and warrant different questions.
- Where a supplier’s external footprint is small, pair observable findings with direct assessment and security documentation.
Where Black Kite helps
Continuous external visibility into the cybersecurity posture of every supplier in your vendor ecosystem. For suppliers with little observable infrastructure, that external view is a starting point rather than a complete picture, and direct assessment fills the gap.
02 Turn Known Vulnerabilities Into Remediation Actions
of AI companies have a critical unpatched vulnerability. 9.8% carry at least one listed in CISA’s Known Exploited Vulnerabilities catalog.
- Start with KEV-listed vulnerabilities. These have a documented history of exploitation, which makes them the easiest findings to prioritize and the hardest for a supplier to deprioritize.
- Identify the affected assets and determine whether they support services your business depends on.
- Ask the supplier for a remediation date, not an acknowledgment.
- Track each finding to confirmed remediation rather than to first response.
Where Black Kite helps
Continuous monitoring and FocusTags® identify which suppliers are exposed to specific vulnerabilities and emerging threats. Through The Bridge™, you can raise a finding with the supplier directly and follow remediation to close.
03 Prioritize Suppliers by Ransomware Exposure and Business Impact
of AI companies reach or exceed the critical ransomware susceptibility threshold
- Measure each AI supplier against the critical threshold to identify the group that warrants closer attention.
- Cross that list with business context: which suppliers support essential operations, and which process sensitive information.
- Escalate the overlap first. A supplier with elevated susceptibility and a critical business role is a different problem from one with the same security profile and limited impact.
- Reassess after remediation. Susceptibility changes as a supplier’s exposure changes.
Where Black Kite helps
The Ransomware Susceptibility Index® evaluates the weaknesses commonly associated with ransomware attacks, including vulnerable internet-facing systems, exposed remote-access services, and compromised credentials. Because it updates continuously, the priority list reflects a supplier’s current posture rather than its posture on the day it was last assessed.
04 Investigate Suspicious Activity and Stolen Credentials
of AI companies showed unusual network behavior. 6.5% had Botnet Infection findings. Corporate login information was associated with 6.8%.
- Treat these findings as questions for the supplier, not conclusions about it. An IP reputation match does not confirm a compromise.
- Ask four things: has the activity been investigated, were affected credentials rotated, were customer-facing services involved, and what changed afterward.
- Prioritize suppliers carrying more than one signal. Among the 1,771 AI companies with a known exploited vulnerability, 14.7% also had Botnet Infection findings.
- For credential findings, establish whether the exposed logins relate to systems that support the service you buy, including cloud platforms, development environments, and collaboration tools.
Where Black Kite helps
Continuous monitoring and threat intelligence surface these signals across your monitored suppliers. When a finding appears, that evidence is what initiates a targeted review rather than a scheduled one.
Layer 2: The components
05 Identify Shared Dependencies Behind AI Products
AI companies carried security findings tied to n8n, Langflow, or LiteLLM.
- Ask every AI supplier which components its product depends on: models, software libraries, workflow platforms, and cloud infrastructure.
- Map those components across your suppliers to find where several depend on the same thing. Suppliers that look independent in a vendor register often are not one level down.
- Ask how the supplier learns that a component has been compromised, and how quickly it can respond when one is.
- Manage shared components as concentration risk, the same way you already manage several critical vendors running on one cloud platform.
Where Black Kite helps
FocusTags provide vendor-specific intelligence about significant technology vulnerabilities and threat campaigns. Supply chain visibility extends that view into supplier relationships and shared dependencies, so the components sitting beneath several suppliers become visible before an incident makes them visible.
Layer 3: The integrations
06 Understand and Limit What AI Tools Can Access
of recorded MCP security findings were rated High or Critical. Among Critical findings, 87.0% involved code or command injection.
- Inventory every system each AI product connects to, what information it can reach, and what actions it can perform.
- Apply least privilege to integration permissions, and grant standing access only where the product cannot work without it.
- Ask whether the product runs MCP servers, how those connections are authenticated, and what they can reach.
- Establish how quickly access can be revoked, and confirm the procedure works before you need it.
- Review integration permissions on a schedule. Access granted at onboarding tends to outlive the reason it was granted.
Where Black Kite helps
External cybersecurity intelligence identifies observable exposures associated with your suppliers and their publicly accessible services. The permissions granted to integrations inside your own environment sit outside any external view, so those remain yours to review and control directly.
It’s Time to Manage AI Supply Chain Risk
AI innovation should not come at the expense of supply chain visibility. Understand the suppliers, technologies, and connections behind AI to manage the risks they introduce.
Black Kite shows you which ones they are, what they’re exposed to, and when that changes.
NEXT: Discover the methodology of this report.