Black Kite Research Reveals Widespread Security Gaps Across AI Supply Chain, Exposing Enterprises to Growing Third-Party Cyber Risk

New study of 18,160 AI companies finds one-third have critical unpatched vulnerabilities, 61.7% have DMARC email authentication weaknesses, and nearly 30% reach critical ransomware susceptibility levels

BOSTON, Nov. 10, 2026 — Black Kite, a leader in third-party cyber risk intelligence, today released its 2026 AI Supply Chain Report: The Risks Behind Your AI Relationships, revealing widespread cybersecurity weaknesses across the rapidly expanding ecosystem of companies developing, deploying, and supporting artificial intelligence. The study of 18,160 AI companies worldwide found that despite the sophistication of the technologies they deliver, many AI suppliers continue to struggle with basic security practices, including patch management, email authentication, credential protection, and network security. As organizations increasingly integrate AI into critical business processes, those weaknesses can create new pathways for attackers to reach sensitive data, cloud infrastructure and enterprise systems.

"The AI supply chain isn't failing in exotic ways. It's failing in ordinary ones," said Dr. Ferhat Dikbiyik, Chief Research & Intelligence Officer at Black Kite. "We're seeing companies building incredibly sophisticated AI technologies while leaving critical vulnerabilities unpatched, failing to enforce basic email security, and exposing credentials that could provide access to sensitive systems. These are familiar security problems, but AI is dramatically changing their potential impact. As enterprises give AI suppliers greater access to their data, applications and infrastructure, they're also inheriting the security weaknesses of those suppliers."

The research also maps the AI supplier ecosystem, classifying companies into six categories based on their primary business activities. While 10% provide foundational AI technologies, including models, computing infrastructure and data services, 31% develop AI applications for everyday business operations and another 23% build and implement AI solutions for other organizations. These application and implementation companies are the suppliers most enterprises actually contract with.

Critical Vulnerabilities and Basic Security Failures Are Widespread Across AI Suppliers

Key findings demonstrate that AI innovation does not necessarily translate into cybersecurity maturity:

• One in three AI companies has a critical unpatched vulnerability. Black Kite identified critical vulnerabilities at 34.2% of the companies assessed. Nearly one in ten (9.8%) had at least one vulnerability listed in CISA's Known Exploited Vulnerabilities catalog, meaning attackers have already exploited the underlying flaw in real-world attacks.
• Nearly two-thirds have weaknesses in basic email security. DMARC weaknesses were identified at 61.7% of AI companies, while 37.3% had DKIM weaknesses and 11.2% had SPF weaknesses. These gaps can make it easier for attackers to impersonate legitimate company domains and conduct phishing or fraud campaigns.
• Nearly 30% demonstrate critical ransomware susceptibility. Across the AI ecosystem, 29.7% of companies scored at or above 0.4 on Black Kite's Ransomware Susceptibility Index® (RSI®), indicating elevated exposure to conditions associated with ransomware attacks.
• Suspicious activity extends beyond software vulnerabilities. Black Kite identified unusual network activity at 44.3% of the 18,160 AI companies assessed and botnet-related findings at 6.5%. These are externally observed signals that warrant investigation, not confirmation of a compromise.
• Even leading AI companies have significant security exposures. Among the Forbes AI 50, 52% had critical unpatched vulnerabilities, compared with 34.2% of the broader AI ecosystem. Additionally, 26% had known exploited vulnerabilities and 28% had credentials in infostealer records. Although larger digital footprints can create more opportunities for security findings to be detected, the results reinforce that even the most prominent AI companies require independent cybersecurity assessment.
• Newly disclosed PHP2Shell vulnerabilities highlight shared software exposure. Black Kite identified PHP versions affected by the newly disclosed PHP2Shell vulnerabilities at 939 AI companies (5.2%) in the study. The findings illustrate how vulnerabilities in a single widely used software component can create widespread exposure across the AI ecosystem.

AI Suppliers With the Deepest Customer Access Face Disproportionate Credential Exposure

Black Kite found that companies helping enterprises develop, implement and operate AI technologies experienced higher rates of stolen credential exposure than many other AI categories. Specifically, 19.9% of cloud AI solution providers had corporate login information appearing in infostealer records, compared with just 4.1% of model and platform developers. Data and labeling companies followed at 13.9%, while AI customer service platforms reached 11.5%.

These suppliers often have access to sensitive customer information, development environments, cloud infrastructure and enterprise applications. If stolen credentials remain valid and are exploited, attackers could potentially use them to reach systems supporting the AI services their customers rely on.

Behind each AI supplier is an ecosystem of software components, open-source tools, cloud services and integrations that can introduce additional security exposure.

Black Kite identified 105 AI companies with security findings tied to three widely used AI development and workflow technologies: n8n, Langflow and LiteLLM. These components can connect AI applications to other systems, often handling credentials, workflows or access to external services.

The report examines the March 2026 LiteLLM supply chain compromise, in which malicious software releases introduced credential-harvesting code into downstream environments, including that of AI data services provider Mercor. The incident illustrates how a compromised software component can create risks for organizations that never directly selected or deployed it.

The growing use of AI integrations introduces another layer of risk. When analyzing publicly discoverable Model Context Protocol (MCP) components, which enable AI applications to interact with external tools and business systems, Black Kite identified 1,732 security findings. Nearly half (48.8%) were classified as High or Critical severity. Among the 77 Critical findings, 87% involved code or command injection vulnerabilities.

Rethinking Third-Party Cyber Risk for the AI Era

Black Kite recommends that organizations continuously monitor AI suppliers for security exposures, identify shared software dependencies, prioritize remediation based on business impact, and assess the access granted to AI applications and integrations.

The full 2026 AI Supply Chain Report: The Risks Behind Your AI Relationships is available at https://blackkite.com/reports/ai-supply-chain-report-2026/

Media Contact:
Michelle Kearney
Hi-Touch PR
443-857-9468
[email protected]

SOURCE Black Kite