ContentLora: AI safety and alignmentNew and updated AI safety and alignment pages on ContentLora.https://contentlora.com/enAI controlhttps://contentlora.com/wiki/ai-control/https://contentlora.com/wiki/ai-control/AI control designs safeguards that hold even if an AI model tries to subvert them: monitoring, trusted editing and sandboxing.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaiCAISI / CAISSI: the US government's AI testing centre at NISThttps://contentlora.com/wiki/caisi/https://contentlora.com/wiki/caisi/The US centre at NIST that tests frontier AI models, from AI Safety Institute to CAISI (2025) and CAISSI (2026).Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetypolicyaiConstitutional AIhttps://contentlora.com/wiki/constitutional-ai/https://contentlora.com/wiki/constitutional-ai/Constitutional AI trains models with written principles and AI feedback instead of only human labels. How it works and what changed in 2026.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaiFrontier safety frameworks (responsible scaling policies)https://contentlora.com/wiki/frontier-safety-frameworks/https://contentlora.com/wiki/frontier-safety-frameworks/Frontier safety frameworks are AI companies' if-then rules for dangerous capabilities. How they work, who has one, and 2026 changes.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaipolicyReinforcement learning from human feedback (RLHF)https://contentlora.com/wiki/rlhf/https://contentlora.com/wiki/rlhf/RLHF trains AI models on human preference rankings. Where it came from, why it became the standard, and its limits.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaiScalable oversighthttps://contentlora.com/wiki/scalable-oversight/https://contentlora.com/wiki/scalable-oversight/Scalable oversight asks how humans can supervise AI that outperforms them. Debate, AI-assisted judging and weak-to-strong research.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaiSparse autoencoders (interpretability)https://contentlora.com/wiki/sparse-autoencoders/https://contentlora.com/wiki/sparse-autoencoders/Sparse autoencoders split an AI model's internal activity into readable features. Their rise, scale-up and limits.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaiUK AI Security Institute (AISI)https://contentlora.com/wiki/uk-ai-security-institute/https://contentlora.com/wiki/uk-ai-security-institute/The UK AI Security Institute tests frontier AI models for national-security risks. Its history, tools and key findings.Sat, 10 Oct 2026 00:00:00 GMTWikiai-safetyaipolicyAI safety tracker: alignment, interpretability and evals in 2026https://contentlora.com/events/ai-safety-tracker/https://contentlora.com/events/ai-safety-tracker/Live tracker of AI safety milestones: interpretability results, evaluations, safety frameworks, incidents and institutes.Sat, 10 Oct 2026 00:00:00 GMTDevelopingai-safetyaipolicyCan AI safety keep pace with AI capabilities?https://contentlora.com/analysis/ai-safety-keeping-pace-debate/https://contentlora.com/analysis/ai-safety-keeping-pace-debate/The central debate in AI safety in 2026: are evaluations, interpretability and oversight keeping up with fast-rising capabilities?Sat, 10 Oct 2026 00:00:00 GMTAnalysisai-safetyaiAI safety and alignment in 2026: a crash coursehttps://contentlora.com/explain/ai-safety/https://contentlora.com/explain/ai-safety/A sourced crash course on AI safety: alignment, interpretability, evaluations, oversight, safety institutes and the 2026 frontier.Sat, 10 Oct 2026 00:00:00 GMTExplainerai-safetyaipolicyHow AI safety testing works: evals, red teams and thresholdshttps://contentlora.com/explain/how-ai-safety-testing-works/https://contentlora.com/explain/how-ai-safety-testing-works/How frontier AI models are tested before release: dangerous-capability evals, jailbreak red-teaming, and why testing got harder.Sat, 10 Oct 2026 00:00:00 GMTExplainerai-safetyaiHow mechanistic interpretability works: looking inside AIhttps://contentlora.com/explain/how-mechanistic-interpretability-works/https://contentlora.com/explain/how-mechanistic-interpretability-works/How researchers reverse-engineer AI models: features, sparse autoencoders, circuit tracing and the 2026 Jacobian lens.Sat, 10 Oct 2026 00:00:00 GMTExplainerai-safetyaiWhat is AI alignment? Making AI do what we intendhttps://contentlora.com/explain/what-is-ai-alignment/https://contentlora.com/explain/what-is-ai-alignment/AI alignment explained: how labs train models to follow human intent, why it can fail, and what 2024–2026 studies found.Sat, 10 Oct 2026 00:00:00 GMTExplainerai-safetyai