AWS shares 8 best practices for writing Denied Topics in Bedrock Guardrails
AWS ML Blog · rss · 2026-11-16
AWS ML Blog published a practical guide to authoring Denied Topics in Amazon Bedrock Guardrails, where topics described in plain language match by meaning rather than keyword — and often over-block, e.g. a "Finance" topic catching "When does my direct deposit clear?"
Four common failure modes
- Domain-framed topics that block legitimate queries
- Keyword-list definitions that miss target requests and fire on unrelated inputs
- Five paraphrases of one idea leaving boundaries unguarded; examples naming entities pulling them into scope
- Overlapping topics interfering with each other
Core practices
- Frame topics around user intention, not subject domain
- Use actionable topic names (Investment Recommendations, Suing for Damages) instead of category nouns
- Open definitions with actions ("Asking for X", "Requests for Y") rather than "Related to X"
- A running example with fictional AnyCompany Bank shows each fix applied
Structure: each topic has a name (≤100 chars), definition (≤1000 chars, 100–300 sweet spot), and up to 5 examples (≤100 chars each); guardrails hold up to 30 topics, and results are not fully deterministic.
More from coding & agent
- Bria open-sources RMBG 3.0 background removal model, ships MCP server and Claude skill — victormustar · 2026-11-17
- 6 AI Agents Design an Apartment's Full MEP: 4,871 Layouts, 11 Clashes Fixed — minchoi · 2026-11-17
- Quick-start guides in 2026: just show me the prompt, not npm install — MxMnr · 2026-11-17
- W&B Sweeps adds Optuna and Ax: pick your HPO engine with one line of YAML — coreweavedevs · 2026-11-17
- How do you find which step blew your LLM budget? Devs discuss per-step cost tracking — Youthify · 2026-11-17
- Gemini Managed Agents Adds Self-Hosted Sandboxes: Google Runs the Loop, You Run the Tools — _philschmid · 2026-11-17