Compliance, security enrichment, and smarter autoscaling dominate the week

Federal compliance reaches a new level

Two Datadog releases this week address the growing compliance requirements facing public-sector organizations. The Datadog guide to OMB M-26-14 walks through how the new federal logging framework—with its risk-based approach, 6-month searchable retention requirements, and CISA maturity milestones—maps to Datadog's Flex Logs, Cloud SIEM, and OpenTelemetry collection. On the authorization front, Datadog for Government achieved GovRAMP High certification, adding NIST 800-53 Rev. 5 controls for state and local governments who previously had to lean on the federal-only FedRAMP path.

Smarter observability pipelines

Datadog rolled out two features that help teams reduce costs and enrich data before it reaches analysis tools. The Observability Pipelines CDN cost reduction guide shows how to route high-volume Cloudflare and Akamai edge logs to low-cost object storage while sending only key metrics to Datadog, then query archived logs via Archive Search without indexing every event. Separately, MITRE ATT&CK Enrichment Packs for Observability Pipelines automatically tag security events from Okta, Palo Alto, FortiGate, AWS WAF, CloudTrail, and Windows with relevant tactics and techniques before logs ever reach a SIEM—turning vendor-specific raw logs into analysis-ready telemetry.

AI instrumentation becomes the default

Sentry's JavaScript SDK 10.61.0 enables AI span streaming by default: all gen_ai spans are now extracted from transactions and sent as v2 envelope items, preventing them from being dropped by transaction size limits. AI message data is no longer truncated by default either—self-hosted Sentry users should set streamGenAiSpans: false if they don't want this behavior. The same release also adds setAttribute/setAttributes APIs at the top level, Cloudflare D1 and SQLite Durable Object instrumentation, and a react-router v8 peer dependency bump. Meanwhile, Sentry's log export feature in Explore → Logs now supports CSV and JSON Lines formats, letting you export up to 10,000 log lines with all columns included.

Sentry JavaScript 10.62.0 followed shortly after, adding v7 support for the Vercel AI integration (Cloudflare not yet supported in v7) and fixing a node runtime crash when tracingChannel is unavailable.

Autoscaling, migrations, and security

Datadog shared two internal engineering stories this week. One details how Kubernetes Autoscaling eliminated $3 million in annualized idle compute costs by consolidating manual horizontal and vertical scaling into a single multidimensional resource. The other describes how the team migrated Stream Router from FoundationDB to a relational schema using AI-assisted refactoring with Claude and Cursor, without disrupting production traffic.

On the security front, Datadog's RUM Browser SDK v7.4.0 fixed a high-severity prototype pollution vulnerability and added profiling bridge mode support. Datadog's Agent 7.80.3 fixed a DPA CPU-limit inconsistency in HA mode. In the tracing libraries, dd-trace-py 4.11.0rc5 fixed async generator crashes on Python 3.11–3.14 and an LLM span timeout bug, while dd-trace-py 4.11.0rc4 resolved forked-process trace duplication and Celery span naming issues. Sentry CLI shipped version 3.6.0 with zstd-compressed snapshot manifests and fixes for the update nagger and renamed project slugs.