# 🎤 Agents IA : quels risques pour les entreprises ? (Kader Seddak, Orange Cyberdefense)

Page: https://stenobird.com/podcast/monde-num-rique-actualit-tech-ia-4060320/agents-ia-quels-risques-pour-les-entreprises-kader-seddak-orange-cyberdefense
Text version: https://stenobird.com/podcast/monde-num-rique-actualit-tech-ia-4060320/agents-ia-quels-risques-pour-les-entreprises-kader-seddak-orange-cyberdefense.md
Podcast: [Monde Numérique (Actu tech & IA)](https://stenobird.com/podcast/monde-num-rique-actualit-tech-ia-4060320)
Published: 2026-04-15T05:00:00+00:00
Episode link: https://www.mondenumerique.info/episode/agents-ia-quels-risques-pour-les-entreprises-kader-seddak-orange-cyberdefense
Audio file: https://audio.audiomeans.fr/file/nTLjBrCvPC/385e78fc-1c70-4df3-bc07-39caa0083a9c.mp3?_=1776083820
Processing state: processed
JSON: https://stenobird.com/v1/public/podcasts/monde-num-rique-actualit-tech-ia-4060320/episodes/agents-ia-quels-risques-pour-les-entreprises-kader-seddak-orange-cyberdefense
Duration seconds: 929

## Resource

The rapid rise of AI agents and generative models introduces critical security vulnerabilities like prompt injection and 'Shadow AI.' Companies must move beyond reactive security to a 'security by design' approach that protects data integrity from the start.

## Highlights
- Main idea: The primary risk of AI adoption is the loss of visibility into complex, rapidly evolving automated systems
- Failure mode: 'Shadow AI' occurs when employees use public LLMs with sensitive corporate data, unintentionally exposing company assets
- Technical threat: Prompt injection allows attackers to bypass safety guardrails by manipulating model instructions through creative phrasing
- Practical takeaway: Implement security by design by integrating data and code protection mechanisms during the initial development phase
- Strategic insight: Securing the supply chain is vital, as third-party providers with weak security can become entry points for AI-driven attacks

## Topics

Artificial Intelligence, Cybersecurity, AI Agents, Prompt Injection, Data Privacy, Shadow IT, Cloud Security, Orange Cyberdefense

## Chapters
- 1:00 — The Shift in Corporate Awareness: Comparing the current AI adoption wave to the cloud era, noting that companies are more aware of data risks this time.
- 2:00 — Core AI Security Risks: An analysis of visibility loss and specific threats like prompt injection and model manipulation.
- 5:20 — The Danger of Shadow AI: Discussing the risks of employees using unauthorized public AI tools and the resulting data exposure.
- 8:40 — Securing AI Agents and Flows: How the rise of autonomous agents increases system complexity and the need for controlled data flows.
- 11:50 — Supply Chain Vulnerabilities: The importance of ensuring that third-party vendors and partners maintain high cybersecurity standards.
- 14:10 — Best Practices for Implementation: Recommendations for integrating security into the AI lifecycle, focusing on data and code protection.

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/monde-num-rique-actualit-tech-ia-4060320/episodes/agents-ia-quels-risques-pour-les-entreprises-kader-seddak-orange-cyberdefense/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/monde-num-rique-actualit-tech-ia-4060320/agents-ia-quels-risques-pour-les-entreprises-kader-seddak-orange-cyberdefense.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.