How Enterprise AI Can Increase Ransomware Risk—and How to Limit It

· Original article ↗

Summary

Enterprise AI can accelerate existing ransomware tactics when attackers compromise identities with broad permissions. The article recommends least privilege, monitoring, human approval for risky actions, and recovery planning.

Key points

  • AI assistants and agents may expose sensitive data or enable unauthorized actions when attackers compromise their identities or delegated permissions.
  • The article says AI can speed up existing attack stages, including reconnaissance, credential abuse, data theft, and extortion, rather than create an entirely new ransomware threat.
  • Recommended controls include inventorying AI applications and integrations, limiting permissions, monitoring AI activity, and requiring approval for high-risk actions.
  • Organizations should be prepared to revoke compromised tokens, disable integrations, suspend AI workflows, and restore operations from tested backups.
  • The Acronis Cyberthreats Report H2 2025 describes GTG-2002 using AI in cyber operations and GLOBAL GROUP using an AI chatbot for ransom negotiations; these examples are presented as operational efficiencies, not new infection methods.
  • Prompt injection is one risk, but the article emphasizes that excessive permissions, insecure integrations, and insufficient oversight can create broader exposure.

Article Details

Defense Focus
Reduce ransomware exposure from enterprise AI by governing AI workflows, limiting delegated permissions, monitoring activity, and requiring approval for high-risk actions.
Detection Methods
  • Inventory approved and unauthorized AI applications, models, and integrations.
  • Monitor AI prompts, interactions, usage, and policy violations.
  • Correlate AI activity with identity events, data access, exports, agent actions, and endpoint, SaaS, and cloud telemetry.
  • Maintain audit trails identifying the user or service account initiating actions, accessed resources, and required approvals.
Data Sources
  • AI application usage, prompts, and interactions
  • Identity events and delegated permissions
  • Data access and export activity
  • AI agent actions
  • Endpoint telemetry
  • SaaS telemetry
  • Cloud telemetry
Platforms
  • SIEM systems
  • XDR systems
  • Secure web gateways
  • CASB
  • DLP
Defensive Actions
  • Apply least privilege to users, AI applications, service accounts, and APIs; regularly review delegated permissions and revoke unused credentials.
  • Restrict AI access to only the systems and information required for each task.
  • Use secure web gateway, CASB, and DLP capabilities to discover AI services, restrict unauthorized tools, and prevent sensitive data transfers.
  • Require human or policy-based authorization for bulk exports, administrative changes, external communications, and code execution.
  • Prepare to revoke compromised tokens, disable affected integrations, and suspend AI workflows when malicious activity is detected.
  • Maintain immutable backups and test recovery procedures.

MITRE ATT&CK

Threat Actors

Vendors

Products

Related Articles