Neuralys
Pick-and-choose AI training

Cybersecurity in the age of agents

Three modules on confidential data, prompt-injection leaks and agent isolation, including review of residual risks.

Course program · 3 modules

SEC_01

Confidential data

Objective

Classify sensitive data and bound agent access and handling.

Topics covered

  • Data classification
  • Agent access
  • Sensitive-data handling

Key learning outcomes

  • Classify a scenario’s data by sensitivity.
  • Define an agent’s permitted access and handling.

SEC_02

Data leaks through prompt injection

Objective

Understand untrusted instructions, exfiltration scenarios, and layered defenses.

Topics covered

  • Untrusted instructions
  • Exfiltration scenarios
  • Layered defenses

Key learning outcomes

  • Spot an untrusted instruction in incoming content.
  • Describe a data-leak scenario.
  • Propose multiple layers of defense.

SEC_03

Proper agent isolation

Objective

Choose execution, network, and tool boundaries, then review residual risks.

Topics covered

  • Execution boundaries
  • Network and tool access
  • Residual risks

Key learning outcomes

  • Choose the access needed for an agent scenario.
  • Describe the selected isolation boundaries.
  • Identify risks that still need review.
Discuss this course →