Datadog expands observability platform to autonomous AI team colleague

Contents

At its annual DASH conference, Datadog announced numerous new features for its observability and security platform. The focus is on the AI agent Bits AI SRE, the AI Guard tool for protecting AI applications, and a new deployment model called Bring Your Own Cloud Logs (BYOC Logs), which allows customers to keep their log data in their storage systems.

As Datadog explains in its DASH-2026 announcement, the innovations are intended to help companies better manage increasingly rapid software development and the growing complexity of AI-driven IT landscapes. According to Chief Product Officer Yanbing Li, the stated goal is for companies not only to build better models but also to create “operational control around these systems.”

CLC 2026: Conference for DevEx and Platform Engineering

The CLC Conference, specializing in Developer Experience (DX) and Platform Engineering, will take place in Mannheim from November 11 to 12, 2026. A special focus will be on how Agentic AI is changing the work of developers, software architects, DevOps, and platform engineers, and how digital sovereignty can be sustainably achieved.

Tickets are now available at an early bird price.

Bits AI SRE: Agentic Approach Instead of Classic AIOps

Datadog's AI agent Bits AI SRE, announced last year, is intended to go far beyond the capabilities of classic AIOps approaches. It aims to differentiate itself from competing offerings from Dynatrace, Splunk, or Elastic, which still rely primarily on rule-based correlation engines and pattern recognition via alerts. Bits AI works like an “agentic team colleague” that continuously leverages the context of all Datadog telemetry. The system independently forms multiple root cause hypotheses, tests them through targeted queries, and classifies them as validated, invalidated, or unclear. The agent accesses metrics, logs, traces, topology data, and linked runbooks – for example, from Confluence – and executes exploratory queries across the entire environment. According to Datadog, Bits AI SRE not only significantly accelerates root cause identification but actually enables autonomous operational processes.

Datadog positions the AI agent as a model-agnostic orchestration layer above large language models. However, the manufacturer does not disclose which foundation models are specifically used. Since integrations with OpenAI and connections to development tools like Anthropic's Claude Code exist in other product areas, collaborations with OpenAI and Anthropic are likely.

Automated Remediation with Configurable Guardrails

Fully automated remediation measures without human approval are possible with the new Bits Agent Builder: Teams can create their own AI agents that automate remediation workflows such as rollbacks, restarts, or feature flag rollouts. Datadog emphasizes that all actions are only performed within customer-defined guardrails – for example, via RBAC, policy engines, audit logging, and mandatory approval by on-call personnel. In security-critical environments, such as the financial sector, Datadog recommends an assistance mode where Bits AI SRE suggests and documents, but the final decision remains with the human.

For integration into existing ITIL or ISO 27001 compliant processes, as are common in larger companies in the DACH region, Datadog offers connections to ServiceNow and Jira. Bits AI SRE acts as the first investigation instance: it receives alerts, creates cases, and generates structured incident reports with root cause, impact, and timeline. However, formal change management processes and documentation obligations remain with the customer; the AI agent is intended as a supplement, not a replacement for existing governance structures.

AI Guard protects AI agents from Prompt Injection

With AI Guard, Datadog is responding to the growing risks associated with AI agents, including hidden, malicious prompt attacks that can cause agents to reveal sensitive information. AI Guard combines telemetry tracing with stateful behavioral analysis to detect and block multi-stage attacks and prompt injection attempts across multiple interactions – attack patterns that would go undetected by stateless prompt-response checks.

The policies can be formulated language-independently, for example, via regex patterns or classifiers that recognize IBANs, email addresses, or customer numbers regardless of the prompt's language. Especially in multilingual corporate environments, where, for example, German-language user texts and English system logs meet in mixed prompts, the detection quality ultimately depends on the language model used, according to Datadog. AI Guard is announced as LLM-agnostic; integration is via SDKs for Python, JavaScript, and Java.

BYOC and EU Data Centers for More Data Sovereignty

With Bring Your Own Cloud Logs (BYOC), Datadog addresses the problem of exponentially growing log data volumes from AI workloads. In this model, the platform is operated in the customer's cloud environment, and data is processed and indexed directly in the company's own object storage without having to move it to a datadog-centric environment. If the chosen cloud provider supports it, DACH companies can thus potentially keep their observability data exclusively in EU data centers or the EU region in Germany, which has existed since 2018. GDPR-relevant aspects such as order processing and access by US providers fundamentally remain, but data localization and BYOC technically pave the way for greater sovereignty.

Among the other announcements made at the DASH conference are an Agent Console that provides central monitoring for AI agents and supports development tools like Claude Code, Cursor, and GitHub Copilot. The Bits Detection module independently recognizes anomalies and automatically initiates investigations, while Agent Evals is used for debugging AI agents, including those created by customers themselves.

(map)