Skip to content
Braincap
← All articles

Enterprise GenAI: how it amplifies ransomware risk, and how to contain it

Illustration of a GenAI system connected to enterprise resources, with ransomware risk

Rapid enterprise GenAI adoption brings productivity, but it also widens the attack surface. The central idea, backed by field data, is simple: AI generally doesn’t invent new attack techniques — it accelerates existing ones. Below are the mechanisms by which it amplifies ransomware risk, and the controls that keep it in check.

How GenAI amplifies the risk

  • It inherits identities and permissions. An AI assistant or agent operates with the rights of whoever uses it. If those credentials are compromised, the attacker directly gains the AI’s capabilities too — including a much faster way to locate sensitive information across the company’s knowledge.
  • It expands the attack surface. AI agents with autonomous permissions are riskier than read-only assistants. Excessive delegated authority becomes an attack accelerator. On top of that, through prompt injection, malicious content hidden in the documents the AI reads can manipulate its behaviour.
  • It speeds up attacker operations. AI lets attackers scale reconnaissance, phishing and code generation. The evidence shows it acting as an operational force multiplier, not as a source of entirely new techniques.

What the data shows

  • Microsoft reported roughly 38 million identity-risk detections per day (Microsoft Digital Defense Report 2025).
  • The Cloud Security Alliance documented large-scale OAuth device-code phishing campaigns targeting Microsoft 365.
  • According to the Acronis Cyberthreats Report H2 2025, several actors already use operational AI: one group for credential harvesting and script debugging; a ransomware group that automated ransom negotiations with a chatbot; a state-sponsored group that used “agentic” AI for reconnaissance and vulnerability research.

How to contain it (six controls)

  1. Inventory and governance: track approved and unauthorized AI applications (shadow AI), each with an owner and a risk classification.
  2. Least privilege: limit AI systems’ permissions to strictly what they need, and audit credentials and access regularly.
  3. Traffic controls: use CASB/DLP to restrict access to AI services and prevent uploads of sensitive data.
  4. Monitoring and auditing: correlate AI activity with SIEM/XDR and keep detailed audit trails.
  5. Containment and recovery: prepare token revocation, workflow suspension, and immutable backups (resistant to deletion/encryption).
  6. Human authorization: require approval for high-risk actions (bulk exports, administrative changes, code execution).

The takeaway

The risk doesn’t come from “AI going rogue”, but from access that’s too broad, monitoring that’s too thin, and recovery that isn’t ready. The same principles that contain classic ransomware — least privilege, visibility, immutable backup, human approval — still hold; GenAI just raises the stakes when they’re applied superficially.

Sources