Training · AI security and governance

Technical course · 1 day (7 to 8 h)

Securing generative AI and AI agents

A single booby-trapped document can be enough to hijack an AI assistant. In one day, your teams learn to spot these attacks, reproduce them on labs and design the defences, without slowing projects down.

Who it is for
Developers, architects, CISOs, CIOs, AI project managers
Goal
Attack, then defend, your LLM applications and AI agents
Duration
1 day (7 to 8 h)
Format
In-house or public, on site or online

Your teams will be able to

  • Identify the threats specific to LLMs and agents (OWASP frameworks)
  • Explain and reproduce the major attacks: prompt injection, data leakage, agent hijacking, MCP servers
  • Design defence in depth: least privilege, human approval, isolation, logging
  • Place the obligations: AI Act and GDPR

The day

Morning
Prompt injection, data leakage, RAG, security of agents and MCP servers.
1.5 h workshop
Online labs on training platforms, then defending an agent with our checklist.
Afternoon
Defence in depth, red teaming, AI Act and incident response.

Included: a kit of 5 templates (usage policy, agent checklist, audit and red teaming grids, incident procedure) and 4 online follow-up sessions.

Course outline

Get the full programme

Objectives, audience, prerequisites, agenda and formats in a 3-page PDF, available as soon as the form is sent. We can then get back to you to adapt the format to your organisation.

The documents are currently in French; sessions are available in English. Prefer to talk directly? Book a free 30-minute call.

Get the programme

Something went wrong. Please try again, or email us at contact@nyxora.studio.

Thank you, here is the programme

We will get back to you shortly. To move faster, book a 30-minute call.