Open-source components, build pipelines, AI models, and AI agents will be the main entry points for supply chain attacks in 2027, according to new predictions from Kaspersky. The forecast, published as part of the annual Kaspersky Security Bulletin, also names credential theft and cyberespionage as attackers' primary goals and warns that dependence on a single AI provider is becoming a supply chain risk in its own right.
Supply chain attacks became routine in 2026, with major incidents occurring almost every month. Attackers reached users through the official websites and update servers of trusted software, including Notepad++, Daemon Tools, and CPU-Z. In March, a cascade of compromises spread from the Trivy vulnerability scanner to Checkmarx tools and the LiteLLM AI gateway. AI infrastructure came under attack as well: researchers found hundreds of malicious skills in an AI agent marketplace, and infostealers began hunting for MCP server configurations and AI agent API keys.
A Kaspersky study published in March 2026 found that nearly one in three companies (31%) had faced a supply chain attack over the previous 12 months, making it the most common threat reported. Yet only 9% of respondents ranked it as the most dangerous.
Based on these trends, Kaspersky experts expect the following developments in 2027:
1. Open source will remain the main entry point. Poisoned packages and self-propagating worms that steal developers' tokens will become standard tooling. The growing volume of AI-generated code that nobody has reviewed adds to the risk.
2. Attacks will move down the stack. Adversaries will increasingly target CI/CD pipelines, container registries, and deployment tools, where a malicious payload sees more and survives longer.
3. Malicious modifications will be found in open AI models. Models with deliberately altered weights could spread like ordinary ones and quietly return incorrect output once built into business processes.
4. AI agents and MCP servers will become carriers of backdoors. An agent with broad permissions could wait for a trigger, such as a specific phrase in an incoming email, and then act in the attacker's interest. Shadow AI widens the exposure.
5. Secrets and AI tokens will be the main loot. Cyberespionage and access harvesting will remain the primary motives, with cloud tokens, API keys, and service credentials as the prize.
6. IT and telecom companies will remain a priority target. Compromising one provider can open the door to many of its customers, and the line between supply chain and trusted relationship attacks continues to blur.
7. AI model availability will become a supply chain risk. Regulatory changes in the countries where the most popular AI models are developed could restrict their use abroad, halting business processes without any attack taking place.
"All security is built on trust. Nobody can verify everything on their own: at some point you have to rely on a compiler, an open-source library, or a cloud provider. Attackers understand this too, which is why supply chain attacks are inevitable," said Sergey Soldatov, head of security operations center at Kaspersky. "AI is now part of that chain. It helps attackers build more sophisticated malware, and it is becoming an entry point itself. Companies need to know what they depend on, what access they have granted, and what they will do when one of those links is compromised."
For the full analysis, head over to the Kaspersky Security Bulletin 2026.
To reduce supply chain risks, Kaspersky recommends that organizations:
- Keep an inventory of dependencies, including open-source components, AI models, MCP servers, and agent skills, and check them against threat intelligence. Kaspersky Open Source Software Threats Data Feed helps track malicious and compromised packages.
- Protect CI/CD like a production environment with least-privilege tokens, short-lived secrets, and full rotation after any incident. Kaspersky Container Security verifies images and configurations from build to runtime, audits RBAC least privilege, reduces noise in secret scanning, and tracks session IDs to spot intercepted tokens.
- Put limits on AI agents: run them in an isolated environment, grant only the access they need, and set a policy for tools that employees install on their own.
- Assess suppliers and write security requirements into contracts. Kaspersky Digital Footprint Intelligence helps spot compromised supplier accounts, leaked data, and unpatched external assets before attackers exploit them.
- Ensure continuous monitoring of infrastructure and network traffic with solutions from the Kaspersky Next (XDR) product line or the Kaspersky Managed Detection and Response service.
- Include supplier compromise in the incident response plan, including rapid isolation of the supplier. If an incident has already occurred, Kaspersky Incident Response can help investigate it and deal with the aftermath.
About Kaspersky Security Bulletin
These predictions are part of Kaspersky Security Bulletin (KSB), an annual series of predictions and analytical reports on key shifts in the world of cybersecurity. Each year, Kaspersky experts review the main events of the past 12 months and outline the trends that will shape the threat landscape in the year ahead. Other KSB pieces are available here.