Black Kite Report Reveals Widespread Security Gaps in AI Supply Chains

icon币界网
Share
AI summary iconSummary
A new report from Black Kite, covered in AI + crypto news, reveals significant security gaps in AI supply chains. The study analyzed 18,160 global AI companies and found widespread issues, including unpatched vulnerabilities and weak email authentication. Over 34% had critical unpatched flaws, and 61.7% exhibited DMARC weaknesses. A security breach risk was also identified, with 29.7% vulnerable to ransomware. The report further found that 5.2% of companies were affected by the PHP2Shell vulnerability.
CoinDesk reports:

Boston, November 10 (Reuters) / PR Newswire — Third-party cyber risk intelligence firm Black Kite today released the "2026 AI Supply Chain Report: Risks Behind AI Relationships," revealing widespread cybersecurity vulnerabilities within the rapidly expanding ecosystem of artificial intelligence companies.

This study of 18,160 AI companies worldwide found that, despite the complexity of the technologies they offer, many AI vendors still have shortcomings in fundamental security practices such as patch management, email authentication, credential protection, and cybersecurity. As businesses increasingly integrate AI into critical business processes, these vulnerabilities could provide attackers with new pathways to access sensitive data, cloud infrastructure, and enterprise systems.

Dr. Ferhat Dikbiyik, Chief Research and Intelligence Officer at Black Kite, said: “AI supply chains aren’t failing in some exotic way—they’re failing in ordinary ways. We’re seeing companies build extremely complex AI technologies without patching critical vulnerabilities, implementing basic email security measures, or exposing credentials that could grant access to sensitive systems. These are well-known security issues, but AI is significantly altering their potential impact. As enterprises grant AI vendors greater access to their data, applications, and infrastructure, they’re also inheriting the vendors’ security weaknesses.”

The study also mapped the AI vendor ecosystem, categorizing companies into six groups based on their primary business activities. Although 10% of companies provide foundational AI technologies—including models, computing infrastructure, and data services—and 31% develop AI applications for everyday business operations, another 23% build and implement AI solutions for other organizations. For most enterprises, the companies they actually partner with are typically those focused on applications and implementation.

Critical vulnerabilities and fundamental security oversights are widespread among AI vendors.

Key findings show that AI innovation does not necessarily mean cybersecurity maturity:

  • One-third of AI companies have critical unpatched vulnerabilities. Black Kite identified critical vulnerabilities in 34.2% of the evaluated companies. Nearly one in ten (9.8%) had at least one vulnerability listed in the U.S. Cybersecurity and Infrastructure Security Agency’s (CISA) Known Exploited Vulnerabilities catalog, meaning attackers have already exploited these underlying flaws in real-world attacks.
  • Nearly two-thirds of companies have weaknesses in basic email security. 61.7% of AI companies have DMARC weaknesses, 37.3% have DKIM weaknesses, and 11.2% have SPF weaknesses. These gaps can make it easier for attackers to impersonate legitimate company domains and carry out phishing or fraudulent activities.
  • Nearly 30% of companies exhibit high ransomware susceptibility. Across the AI ecosystem, 29.7% of companies scored 0.4 or higher on Black Kite’s Ransomware Susceptibility Index (RSI®), indicating a high level of exposure to conditions associated with ransomware attacks.
  • Suspicious activity is not limited to software vulnerabilities. Black Kite found that 44.3% of the 18,160 AI companies evaluated exhibited anomalous network activity, and 6.5% showed signs related to botnets. These are externally observed indicators warranting further investigation but do not equate to confirmed compromises.
  • Even leading AI companies exhibit significant security exposures. Among companies on Forbes’ AI 50 list, 52% have critical unpatched vulnerabilities, higher than the 34.2% in the broader AI ecosystem. Additionally, 26% have known exploited vulnerabilities, and 28% have had credentials recorded in information-stealing malware logs. Although a larger digital footprint may increase the likelihood of discovering security issues, these results indicate that even the most prominent AI companies require independent cybersecurity assessments.
  • The newly disclosed PHP2Shell vulnerability highlights the risks of shared software exposure. Black Kite’s research found that 939 AI companies (5.2%) are using PHP versions affected by the newly disclosed PHP2Shell vulnerability. This finding demonstrates how a vulnerability in a single, widely used software component can lead to widespread exposure across the entire AI ecosystem.

The AI vendor with the deepest access to customer permissions faces a higher risk of credential exposure.

Black Kite found that companies helping businesses develop, implement, and operate AI technologies have a higher rate of compromised credentials compared to many other AI categories. Specifically, 19.9% of cloud AI solution providers had their enterprise login credentials exposed in data-stealing program logs, compared to just 4.1% for model and platform developers. Data and annotation companies had a rate of 13.9%, and AI customer service platforms had a rate of 11.5%.

These vendors typically have access to sensitive customer information, development environments, cloud infrastructure, and enterprise applications. If stolen credentials remain valid and are exploited, attackers could gain access to the systems underlying the AI services that customers rely on.

Behind every AI vendor lies an ecosystem of software components, open-source tools, cloud services, and integrations that may introduce additional security exposures.

Black Kite found that security findings across 105 AI companies are linked to three widely used AI development and workflow technologies: n8n, Langflow, and LiteLLM. These components enable AI applications to connect to other systems and often handle credentials, workflows, or access to external services.

The report also analyzes the LiteLLM supply chain compromise in March 2026, during which malicious software introduced credential-stealing code into downstream environments, including that of the AI data service provider Mercor. This incident illustrates how a compromised software component can pose risks to organizations that never directly selected or deployed it.

The increased use of AI integration has introduced another layer of risk. In analyzing publicly discoverable Model Context Protocol (MCP) components, Black Kite identified 1,732 security findings. Nearly half (48.8%) were classified as high or critical severity. Of the 77 critical findings, 87% involved code injection or command injection vulnerabilities.

Reconsidering Third-Party Network Risks in the AI Era

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

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

About Research

The 2026 AI Supply Chain Report, compiled by Black Kite Research Group™, examines the externally observable cybersecurity posture of 18,160 AI companies worldwide. The study evaluates AI vendors for vulnerabilities, email authentication, credential exposure, suspicious network activity, and ransomware susceptibility, while also assessing software dependencies and AI integration. The findings reflect externally observed security conditions and do not necessarily indicate that assessed companies have been confirmed as compromised.

About Black Kite

Black Kite is an AI-native third-party cyber risk platform built for the connected world. By distilling billions of external risk signals from millions of monitored organizations, Black Kite delivers trusted intelligence to power a connected defense network, enabling organizations to identify risks earlier, act faster, and shift from isolated defenses to collective resilience. With Black Kite, organizations gain greater control, earlier warnings, and confidence in securely collaborating with third parties at scale. Black Kite has earned numerous industry awards and customer accolades. For more information, visit www.blackkite.com or the Black Kite blog.

Media contact:

Michelle Kearney

Hi-Touch PR

443-857-9468

[email protected]

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.