Blocklist

In the Ultra Transcenders books

AI-300AI-103

Each book explains Blocklist in context, with comparison tables and the common traps.

Terms in this definition

• Content filter

  Guardrail set up in Foundry that rates prompts and completions for risks such as hate, violence, sexual content, self-harm, prompt attacks and protected material, blocking anything at or above a chosen severity.

• Stop sequence

  One of up to four strings that make the model halt generation; the sequence itself is not included in the output.

Related terms

• AnalyzeTextOptions

  Request class in the Content Safety SDK carrying the text, categories, blocklist names and output type; AnalyzeText accepts it.

• API Management llm-content-safety policy

  Policy for the AI gateway in API Management: it passes prompts or completions to Azure AI Content Safety and rejects the call with 403 if a blocklist matches, Prompt Shields detect an attack, or a harm category passes its threshold, where 0 is strictest and 7 most lenient.

• BlocklistClient

  Client in the Content Safety SDK for managing text blocklists: creating, listing and deleting them, and adding or removing items (add_or_update_blocklist_items, remove_blocklist_items with at most 100 per call, delete_text_blocklist).

• Custom categories (Content Safety)

  Lets Content Safety detect categories you define, either in standard form (trained over several hours and retiring on 1 September 2026) or in rapid form from samples; for a handful of niche terms, a blocklist takes less effort.

• haltOnBlocklistHit

  A flag on the Analyze text request. When true, analysis of harm categories stops as soon as a blocklist term matches; when false, every analysis runs anyway.

• Monitor online activity (Content Safety Studio)

  Reporting page in Content Safety Studio that summarises how the moderation API has been used, with trends in category and severity, block rate, latency and blocklist matches. No new content is tested here.

• TextBlocklistMatch

  Returned in blocklists_match, this Content Safety result type reports a blocklist hit: the blocklist's name, the item ID and the text that matched.