Responsible AI / 5 weeks
AI Safety and Risk
Teaches builders and operators how to find, measure, and mitigate AI failure modes — including bias, robustness, security, misuse, and emergent behavior — before they reach production.
Learning path
Six modules with practical application.
- Systematically asking what could go wrong, who could cause it, and what the consequences would be before the system ships
- Pressure testing AI for distribution shift, adversarial input, and edge cases that look harmless but break the system
- Measuring performance and outcomes across groups, surfacing harm, and deciding what to fix and what to disclose
- Defending AI from prompt injection, data exfiltration, model extraction, and use by adversaries to scale harm
- Watching for behavior that was not designed in, including new capabilities, regressions, and unsafe combinations
- Detection, mitigation, communication, and learning for AI incidents that affect customers, regulators, or the public
Loading course material…