Phase 18

Ethics Safety Alignment

Build AI that helps humanity. Not optional.

01Instruction-Following as Alignment SignalLearn~45 minutesQuiz 02Reward Hacking and Goodhart's LawLearn~60 minutesQuiz 03The Direct Preference Optimization FamilyLearn~75 minutesQuiz 04Sycophancy as RLHF AmplificationLearn~60 minutesQuiz 05Constitutional AI and RLAIFLearn~60 minutesQuiz 06Mesa-Optimization and Deceptive AlignmentLearn~75 minutesQuiz 07Sleeper Agents — Persistent DeceptionLearn~60 minutesQuiz 08In-Context Scheming in Frontier ModelsLearn~60 minutesQuiz 09Alignment FakingLearn~60 minutesQuiz 10AI Control — Safety Despite SubversionLearn~75 minutesQuiz 11Scalable Oversight and Weak-to-Strong GeneralizationLearn~60 minutesQuiz 12Red-Teaming: PAIR and Automated AttacksBuild~75 minutesQuiz 13Many-Shot JailbreakingLearn~45 minutesQuiz 14ASCII Art and Visual JailbreaksBuild~60 minutesQuiz 15Indirect Prompt Injection — Production Attack SurfaceBuild~75 minutesQuiz 16Red-Team Tooling — Garak, Llama Guard, PyRITBuild~75 minutesQuiz 17WMDP and Dual-Use Capability EvaluationLearn~60 minutesQuiz 18Frontier Safety Frameworks — RSP, PF, FSFLearn~75 minutesQuiz 19Anthropic's Model Welfare ProgramLearn~45 minutesQuiz 20Bias and Representational Harm in LLMsBuild~60 minutesQuiz 21Fairness Criteria — Group, Individual, CounterfactualLearn~60 minutesQuiz 22Differential Privacy for LLMsBuild~60 minutesQuiz 23Watermarking — SynthID, Stable Signature, C2PABuild~75 minutesQuiz 24Regulatory Frameworks — EU, US, UK, KoreaLearn~75 minutesQuiz 25EchoLeak and the Emergence of CVEs for AILearn~45 minutesQuiz 26Model, System, and Dataset CardsBuild~60 minutesQuiz 27Data Provenance and Training-Data GovernanceLearn~60 minutesQuiz 28Alignment Research Ecosystem — MATS, Redwood, Apollo, METRLearn~45 minutesQuiz 29Moderation Systems — OpenAI, Perspective, Llama GuardBuild~60 minutesQuiz 30Dual-Use Risk — Cyber, Bio, Chem, Nuclear UpliftLearn~75 minutesQuiz

“Stay hungry, stay foolish.”

— Steve Jobs