Course 02 · Practical security for AI systems
AI Security
Nine video modules on protecting AI systems: data poisoning, adversarial attacks, prompt injection, model theft, privacy, supply chains, production security, and responsible governance.
AI Security Modules
Follow the nine modules in order, or choose the topic most relevant to the AI systems you work with.
What is AI Security?
Identify the models, data, prompts, credentials, and system boundaries that AI security needs to protect.
Watch lessonData Poisoning & Training Data Risks
Understand how compromised training data and hidden backdoors can influence model behavior.
Watch lessonAdversarial Attacks
Explore evasion attacks, attacker capabilities, and layers of defense against adversarial inputs.
Watch lessonPrompt Injection & LLM Security
Understand how untrusted prompts and retrieved content can cross application boundaries, and how to constrain their effects.
Watch lessonModel Extraction & IP Protection
Examine model theft, extraction risks, and access controls for protecting AI intellectual property.
Watch lessonPrivacy Risks
Explore membership inference, data memorization, and controls for protecting sensitive information in AI systems.
Watch lessonAI Supply Chain Security
Examine risks in model artifacts and dependencies, including unsafe serialization, provenance, and release controls.
Watch lessonSecuring AI in Production
Connect resource limits, validation, monitoring, and action boundaries with safer AI operations.
Watch lessonGovernance, Compliance and Responsible AI
Connect responsible AI practices, governance, and compliance considerations with clear ownership and operational evidence.
Watch lesson