AI Trust & Safety for PMs: Certification in Security, Privacy & Governance

Godhuli Chatterjee

Closed session

This course is popular

3 people enrolled last week.

The One AI PM Skill That Doesn't Expire When the Architecture Does - Security

RAG gave way to MCP. A2A and agent registries are already reshaping how agents talk to each other. JEPA-style architectures are challenging the transformer's dominance. Every layer of the AI stack is being rewritten every 12–18 months; except one. Whoever is building, buying, or regulating AI still needs someone who can answer "is this safe to ship, and can we prove it." That skill doesn't expire when the architecture changes. It's the one PM competency built to survive the next decade of churn, from ANI to AGI.

What you’ll learn

By the end of this course, trust and safety should stop feeling like a separate discipline you consult on the side and start feeling like

Fundamentals: jailbreaking, data and cache poisoning, model theft, supply chain risk, excessive agency, insecure output handling
Privacy and Reliability

hallucination, bias and fairness, toxic and harmful content generation, misuse potential
Governance and Accountability

Workshop agenda

Trust and Safety
the design choices that are unique to AI products: calibrated confidence signaling, explainability, correctability, appropriate friction before high-stakes actions, and expectation-setting
Governance and Accountability
The organizational layer that ties everything else together: human-in-the- loop design, audit trails, and regulatory compliance.
Privacy and Security
What happens to a users data: PII guardrails, data retention and minimization, consent, and the right to deletion and explanation.