Topics
Browse posts by category and tag — every topic we cover, with the latest pieces under each.
Tags
- #content-moderation 11
- #llm-safety 11
- #guardrails 5
- #production 5
- #prompt-injection 5
- #llm-security 3
- #owasp-llm-top-10 3
- #rag 3
- #ai-safety 2
- #api-review 2
- #classifier 2
- #factuality 2
- #hallucination 2
- #image-moderation 2
- #jailbreak-detection 2
- #llama-guard 2
- #meta 2
- #openai-moderation 2
- #output-validation 2
- #safety-classifier 2
- #accuracy 1
- #ai-security 1
- #architecture 1
- #benchmark 1
- #conversation-control 1
- #cost 1
- #ensemble 1
- #false-positives 1
- #google-jigsaw 1
- #llm-firewall 1
- #multimodal 1
- #nemo-guardrails 1
- #nvidia 1
- #omni-moderation 1
- #ops 1
- #perspective-api 1
- #retrieval-augmented-generation 1
- #runtime-filter 1
- #text-moderation 1
- #thresholds 1
- #toxicity-detection 1
- #trust-and-safety 1
- #user-experience 1
- #video-moderation 1
Categories
Tool Comparisons 6 posts
- Prompt Injection Detection Tools: A Practitioner's EvaluationLakera Guard, Protect AI's DeBERTa classifier, Meta Prompt Guard, NeMo Guardrails, and LLM Guard on latency, false positives, and pipeline fit.
- Hallucination Detection Tools Compared: 11 OptionsEleven hallucination detection tools compared by family, deployment, and grounding requirement: HHEM, Lynx, MiniCheck, RAGAS, Bedrock, TLM, and more.
- Best AI Content Moderation Tools 2026: Platform ComparisonA comparison of the leading AI content moderation platforms in 2026, covering Azure AI Content Safety, Hive, AWS Rekognition, and Perspective API.
- Fine-Tuned Classifiers vs. Moderation APIs: Cost and TradeoffsOff-the-shelf moderation APIs are cheap to start and expensive to outgrow. Fine-tuned classifiers are the reverse, and the crossover has a measurable point.
- Image & Video Content Moderation Tools (2026)Text moderation gets the attention, but image and video hold the hard problems. A map of the cloud APIs, self-hosted options, and the CSAM referral path.
- Llama Guard vs Llama Guard 2 vs Llama Guard 3: Lineage ClarifiedMeta's Llama Guard series gets cited loosely, often with the wrong base model or taxonomy. Here is the verified lineage, version by version, with sources.
Tool Reviews 5 posts
- Omni-Moderation API Guide: Categories, Scores, LimitsWhat omni-moderation-latest actually returns: all 13 categories, which ones accept images, how to read category_scores, and where the free endpoint stops.
- Perspective API Review: 8 Attributes and Where It FailsPerspective API scores toxicity, insult, threat and five other attributes for comment moderation, and structurally misses the harms in LLM outputs.
- OpenAI Moderation API Review: What It Catches and MissesOpenAI Moderation API review: ~20ms latency, free with API credits, image support, and the obfuscation, context, and customization gaps it leaves open.
- Llama Guard Benchmark Review: Real Performance vs. Vendor ClaimsMeta's Llama Guard is the default open-source moderation classifier. Its benchmark numbers look strong; here is what they do and do not actually measure.
- NeMo Guardrails in Production: What It Does Well; Where It FailsNVIDIA's NeMo Guardrails offers conversation-flow control that plain classifiers cannot provide. The deployment complexity is real and worth costing out.
Production Practice 4 posts
- How to Mitigate Prompt Injection Attacks: Defense in DepthNo single control stops prompt injection. This guide covers the layered defenses that reduce risk in production, from input hardening to monitoring.
- Content Moderation for RAG: The Retrieval Layer Is an Attack PathRAG pipelines have a moderation gap at the retrieval layer that input and output classifiers miss, because injected documents enter the context first.
- Classifier Ensembles for Production Content ModerationSingle classifiers have characteristic blind spots. Ensembles that mix architectures and training data cut those failures, at a real cost in latency.
- False Positive Costs in Content Moderation: How to Measure ThemFalse positives in content moderation drive hidden costs: user abandonment, review queue spend, and appeal load. How to quantify and calibrate them.
Detection Methods 2 posts
- LLM Hallucination Detection Methods Explained: Four TechniquesThe four main LLM hallucination detection methods compared: sampling consistency, claim decomposition, internal-state probing, and NLI faithfulness checks.
- How to Detect Jailbreak Prompts: A Practitioner's GuideHow to detect jailbreak prompts with perplexity heuristics, trained classifiers, and multi-turn scoring, plus the blind spots of each approach.