ai@7.0.165
Vercel AI SDK 7.0.165 addresses telemetry security and privacy by sanitizing provider request URLs to prevent credential leaks and omitting transcription word timings when recordOutputs is disabled.
Why this signal matters
Vercel AI SDK release 7.0.165 introduces key telemetry and tracing privacy fixes. Provider request URLs are now sanitized prior to text generation results reaching telemetry integrations or tracing subscribers, preventing potential credential exposure embedded in endpoint URLs. Additionally, transcription word timings are removed from telemetry events when the recordOutputs setting is set to false. The patch also bumps internal package dependencies: @ai-sdk/provider-utils@5.0.76, @ai-sdk/provider@4.0.37, and @ai-sdk/gateway@4.0.138.
Actionable summary
ai@7.0.165 implements telemetry data sanitization: provider request URLs are sanitized before reaching telemetry and tracing integrations to protect credentials, and transcription word timings are excluded when recordOutputs is false. Dependency updates include @ai-sdk/provider-utils@5.0.76, @ai-sdk/provider@4.0.37, and @ai-sdk/gateway@4.0.138.
- Agent usefulness
- 96/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Upgrade the ai package to version 7.0.165 or later to ensure endpoint credentials are sanitized in telemetry streams.