Five AI Risks That Can Get You Fired—And How to Avoid Them

Item: Five AI Risks That Can Get You Fired—And How to Avoid Them
Rating: 4
Author: PostSingularity Institute

by Martin Keen

Applied Sciences & Engineering AI & Society UBDigital and Information technology: general topics UBJDigital and information technologies: social and ethical aspects

🎙 Martin Keen 👥 1.8M 📅 May 24, 2026 ⏱ 10 min 👁 87K 📄 expert opinion 🧭 2026-08-06

Available in: English (current) Français

Keywords

shadow AIdata leakagehallucinationprompt injectionAI agents

Summary

In this video, Martin Keen from IBM Technology outlines five AI-related risks that can lead to employee termination, emphasizing the importance of AI governance. The first risk is shadow AI, where employees use unapproved AI tools, potentially causing data breaches. The second is data leakage, where sensitive information is pasted into unauthorized tools, losing control and possibly being used for model training. The third is hallucination laundering, where employees present AI-generated false content as their own work, citing examples of lawyers using fabricated case citations. The fourth is prompt injection, where attackers manipulate AI systems through direct or indirect inputs, overriding instructions. The fifth is unauthorized agentic AI, where employees create autonomous agents that may become ‘zombie’ agents, posing security risks. Keen advises that using AI without governance or verification is dangerous, and suggests that not using AI at all is also risky. He recommends clear policies, approved tools, and verification to mitigate these risks.

155 words

Critical Evaluation

The video provides a clear and accessible overview of five significant AI risks in the workplace, each illustrated with practical examples and potential consequences. Martin Keen, an IBM expert, delivers the content with authority, and the references to the IBM Cost of a Data Breach report lend credibility. The structure is logical, progressing from individual misuse (shadow AI) to systemic vulnerabilities (prompt injection) and emerging threats (unauthorized agents). The argumentation is solid, emphasizing that these risks are not hypothetical but have already caused real damage. However, the video lacks depth in mitigation strategies; while it mentions AI governance, it does not provide specific frameworks or best practices. The sources are limited to IBM’s own resources, which, while relevant, may introduce bias. The title is accurate, and the content aligns well with it. The video is aimed at a broad professional audience, but the analysis here focuses on the merits of the content itself. The presentation is engaging, with clear graphics and a conversational tone, making complex topics accessible. The main weakness is the absence of diverse perspectives or counterarguments, which could strengthen the critical evaluation. Overall, the video is informative and practical, but it could benefit from more detailed guidance on implementing AI governance.

204 words

Title / Content Match

The title accurately reflects the content, listing five specific AI risks and how to avoid them, with a slight clickbait tone that is acknowledged.

Quality & Reliability

8/10

The video is presented by an IBM expert, referencing the IBM Cost of a Data Breach report and providing practical examples. It is well-structured and aligns with industry knowledge, though it lacks detailed citations and is primarily opinion-based.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

AI Governance — Mentioned as a resource for learning more about AI governance.
IBM Newsletter for AI Updates — Mentioned as a way to stay updated on AI news.

Concurring Sources

IBM Cost of a Data Breach Report — Cited in the video to support the prevalence of shadow AI-related breaches.

Contribution & Novelties

The video synthesizes common AI risks into a memorable list, emphasizing the accountability of employees and IT teams. It highlights the emerging threat of ‘zombie AI agents’ and the importance of governance.

Pour aller plus loin :

OWASP Top 10 for LLM Applications — Relevant for understanding prompt injection and other LLM vulnerabilities.
IBM Cost of a Data Breach Report — Provides data on the impact of shadow AI and data breaches.
NIST AI Risk Management Framework — Offers a structured approach to AI governance and risk management.

88 words

Radar Profile

The radar profile shows high scores in information quantity, quality, and reliability, with a slightly lower technical depth, indicating a well-balanced and accessible presentation suitable for a broad professional audience.

Reliability 8/10

💬 Positive: The comments are overwhelmingly positive, with viewers praising the clarity and relevance of the content, and many sharing additional insights or personal experiences related to the risks discussed.