AI & Cybersecurity

Watch and track your favorite playlist.

Curated by: Network Intelligence (66 videos)


Currently Playing: LLM Vulnerabilities Explained: Adversarial Attacks, Jailbreaks & Data Poisoning

Can AI be hacked into lying? Behind every powerful model is a hidden battlefield, where attackers craft prompts, poison data, and manipulate systems at their core. There’s a hidden conflict raging inside today’s most advanced AI systems, a silent war between attackers trying to manipulate models and defenders racing to secure them. In this video, we uncover the dark art of Adversarial Machine Learning (AML) where hackers deceive AI systems into seeing one thing and saying another. Here’s what you’ll learn: • What evasion attacks and adversarial examples really are, and how a few invisible pixels can make an AI misread a stop sign. • How data poisoning secretly plants backdoors into training datasets. • Why prompt injection and jailbreaks can turn your trusted chatbot into something completely different. • How defenders are fighting back with adversarial training, data sanitization, and model alignment. This is more than a technical challenge — it’s an arms race for the future of trustworthy AI. As over 5,000+ papers have already explored, this battle is just beginning. Watch till the end to understand how we can keep AI systems aligned, resilient, and on our side. Watch now and join the discussion: How do you think we can make AI safe from itself? Watch our full AI & Cybersecurity Learning Series: Deep-dive into topics like LLM Security, Red-Teaming AI, Threat Detection, and Responsible AI — all available on our YouTube channel. https://www.youtube.com/watch?v=RQdsmm98sds&list=PLXVUBNOa2d7YyqWr_DgUHw7RwQLE7P24m Timestamps: Chapters below, jump to the topic you’re most interested in! 00:00 – The Hidden War Inside AI 01:02 – What If AI Could Lie? 02:10 – Introduction to Adversarial Machine Learning (AML) 03:25 – How AI Was Tricked: The Evasion Attack 05:00 – Invisible Noise, Real Consequences 06:12 – Data Poisoning: Teaching AI to Fail 08:03 – The Backdoor Trigger Explained 09:15 – Enter the Modern Battlefield: Generative AI 10:22 – Jailbreak Prompts and the DAN Example 11:38 – Prompt Injection: Hacking Through Conversation 13:05 – Defending the Model: The Security Countermeasures 13:58 – Adversarial Training: Vaccinating AI 15:10 – Data Sanitization and Model Alignment 16:22 – The Arms Race: Attackers vs Defenders 17:40 – Why This War Is Just Beginning 18:45 – Over 5,000 Papers and Counting 19:30 – The Core Question: Keeping AI on Our Side 20:15 – Watch Our AI & Cybersecurity Learning Series #AdversarialAI #MachineLearningSecurity #LLM #Cybersecurity #AIAlignment #PromptInjection #DataPoisoning #AIHacking #NetworkIntelligence


Tracks in this Playlist