Course · AI Security: Prompt Injection, Agent Permissions & Data Leak Prevention

AI Security Training for Organisations (Prompt Injection & Data Leaks)

Once an organisation lets AI read email, open files and do work on people's behalf, new routes of attack follow. This course helps IT and security teams understand the attacks that really happen, test their own systems hands-on, and put in place measures that work with the tools the organisation already uses.

Level
Advanced
Length
1 day
Hours
09:00–16:00
Tools
The AI assistants and agents your organisation uses
Format
Onsite or online, hands-on

This course runs in-house at your organisation, taught in English or Thai, with cases and exercises rebuilt from the work of the department taking it.

Request a quoteAll courses

What you will be able to do

  • Explain the main attacks on AI systems, from prompt injection to pulling data out through tools
  • Test your organisation's AI assistants and agents with a basic attack test set
  • Design agent permissions on the least-privilege principle, with the points where a person must approve
  • Put in place measures against data leaks through AI tools, on both the technical and the policy side
  • Write AI security guidelines that development and business teams can share

One-day schedule

09:00–16:00 · topics, cases and exercises can be adapted to your organisation’s work · can be extended to two days for a longer workshop on real work

  1. 09:00–10:30

    The attack surface of AI systems

    • Models, data, tools and people
    • Examples of attacks that have really happened

    ActivityMap the attack surface of the AI assistants in your own organisation

  2. 10:30–10:45

    Break

  3. 10:45–12:00

    Prompt injection and data leaks

    • Direct and indirect prompt injection, and how to test for them
    • Data leaks through conversations, through tools and through an agent's memory

    ActivityTest a sample AI assistant with a basic attack test set

  4. 12:00–13:00

    Lunch

  5. 13:00–14:30

    Agent permissions and protective measures

    • Least-privilege permission scope, and the points where a person must approve
    • Layered defences: on the input, on the tools and on the output

    ActivityRedesign the permissions of one agent, and set its approval points

  6. 14:30–14:45

    Break

  7. 14:45–16:00

    Monitoring, response and guidelines

    • Logging, monitoring and responding when an incident happens
    • AI security guidelines that development and business teams can share

    ActivityWorkshop: test and fix your organisation's AI systems

Frequently asked questions

Who is it for?
IT, security and development teams who look after the organisation's AI assistants or agents. Security leaders can join the overview and measures modules without doing the hands-on testing.
How is it different from the AI Governance course?
AI Governance focuses on policy, risk and oversight at the organisation level. This course is technical: hands-on testing and protecting real systems. Most organisations run both, for different groups.

Contact

Talk to us about training your team

Send a rough brief — how many people, which functions, and what they should be able to do afterwards — and we reply with a format and a quotation. We correspond in English or Thai; quotations and documents are issued by Trainiac Co., Ltd. for your procurement process.

Email
hello@trainiac.ai
Phone
061-894-0404
LINE Official Account
@trainiac
Email us