OpenAI adds inline moderation to completion APIs
OpenAI has integrated moderation scores directly into its Responses and Chat Completions APIs, allowing developers to receive safety signals and confidence scores within a single inference roundtrip. This update eliminates the latency and orchestration overhead of calling the standalone Moderation API separately.
Inline moderation is a massive quality-of-life update that drastically reduces latency and simplifies guardrail implementation, though it won't solve complex jailbreaks on its own.
- –**Latency Optimization**: By returning moderation signals in the same response, developers save a round-trip API call, significantly improving the responsiveness of user-facing chat apps.
- –**Lower Boilerplate**: Eliminating the need to orchestrate standalone moderation requests reduces the complexity of agentic workflows.
- –**Reinforcing the Responses API**: Embedding this in the new Responses API primitive signals OpenAI's direction toward making agent orchestration more stateful and self-contained.
- –**Still Needs Defense-in-Depth**: A simple confidence score list cannot catch sophisticated prompt injections, meaning developers must still maintain secondary safety layers.
DISCOVERED
3h ago
2026-06-04
PUBLISHED
3h ago
2026-06-04
RELEVANCE
AUTHOR
OpenAIDevs