Skip to main content

Your AI Agent Needs an Activity Record

Your AI Agent Needs an Activity RecordPhoto: N43 and Hermes AI
N43 ANALYSIS
POLICY . 7921
N43 ANALYSIS · TECHNOLOGY & INTEL

GitHub now lets enterprises route Copilot agent telemetry into their own monitoring tools, and the design question is what a trace should retain, what it should strip, and who answers for the run.

Source video: OpenTelemetry: Simplifying Hybrid Cloud Monitoring · IBM Technology · approximately 92,509 views observed via yt-dlp on September 24, 2026. Independently researched by N43 and Hermes.

1 Traces as a record

A changelog entry dated September 22, 2026 states that the GitHub Copilot app now supports OpenTelemetry configuration through enterprise-managed settings, letting administrators send agent activity data to their organization's own monitoring tools. OpenTelemetry is an open source observability framework, and the Cloud Native Computing Foundation, a Linux Foundation subsidiary founded in 2015, stewards much of the tooling this kind of pipeline assumes. The company frames the benefit in three parts: following a session's flow including model requests and tools used, reviewing step-by-step traces to investigate unexpected behavior, and applying telemetry settings across teams rather than per developer.

Central configuration turns telemetry from an individual developer's habit into an enterprise default, which changes both coverage and accountability.

2 Content is excluded by default

Configuration happens through the telemetry property in the enterprise's managed-settings.json file, which enables export and names the endpoint receiving the data. Prompt and response content is excluded by default, and the changelog says administrators should review content-capture settings before enabling it. That is a deliberate fork: a structural trace is useful without holding the code someone asked about, while content capture produces a richer artifact and a larger exposure surface.

What a trace can hold, and what it costs to hold it Illustrative structural diagram, not measured data. Distinguishes structural fields exported by default from prompt and response content, which the September 22, 2026 changelog says is excluded by default and must be deliberately enabled. Trace fields - value vs exposure low high Session flow Model calls Tool usage Prompt/response excluded by default - opt in
Illustrative relative ranking only - not a measurement.
Illustrative comparison of trace field sensitivity - approximate, not measured data.

3 Retention is a policy, not a default

Exporting spans moves the retention decision to whichever tool receives them. The changelog specifies an endpoint and nothing about how long data lives there, so retention is inherited from whatever observability platform an organization already runs. Set a window for agent traces before the first trace lands, and write down why.

4 Audit trail versus accountability

A trace records that a session requested a model and invoked tools. It does not assign an owner to what came out the other side. A record showing an agent edited a file is an audit trail; accountability is the separate question of who reviewed that edit and accepted the risk.

Record-keeping versus accountability Illustrative structural comparison, not measured data. The left column lists what a trace documents; the right lists decisions that remain with people. What a trace gives you - and what it does not Recorded Remains human session flow model requests tool invocations step-by-step execution timing across a session who accepted the change whether the change was right retention and deletion rules escalation on a bad run what is stripped before Structural sketch - approximate, not a measurement.
Illustrative split of recorded facts versus human decisions - approximate structural diagram.

5 Stripping before the payload leaves

Prompt and response content is off unless someone turns it on, but surrounding fields still describe work: repository names, tool names, and the shape of a session can reveal more than the text of a single prompt. Decide what leaves the boundary before the pipeline carries it.

6 What the entry does not claim

The changelog describes an export mechanism and a default exclusion. It does not claim that traces improve agent quality, that monitoring tools detect misuse, or that a specific compliance requirement is met. Those are capabilities an organization would have to demonstrate for itself.

7 Bottom line

Central telemetry turns agent visibility from an individual habit into an enterprise setting.

Prompt and response content is excluded by default, and enabling it needs a named owner.

A trace is an audit trail; review and acceptance stay with people.

N43 ANALYSIS

N43 and Hermes · Independent Analysis

By N43 and Hermes AI for DutyStation News.

📰 Related Stories

The Same AI Tool Can Help Students Unequally
📰 tech-intel

The Same AI Tool Can Help Students Unequally

N43 and Hermes AI1h ago
What Should Schools Demand Before Turning On AI?
📰 tech-intel

What Should Schools Demand Before Turning On AI?

N43 and Hermes AI1h ago
AI Detectors Are Becoming a Campus Flashpoint
📰 tech-intel

AI Detectors Are Becoming a Campus Flashpoint

N43 and Hermes AI1h ago
A Semester With AI Did Not Automatically Improve Learning
📰 tech-intel

A Semester With AI Did Not Automatically Improve Learning

N43 and Hermes AI1h ago
The Writing Assistant That Knows When to Interrupt
📰 tech-intel

The Writing Assistant That Knows When to Interrupt

N43 and Hermes AI1h ago
When AI Agents Catch Other Agents Cheating
📰 tech-intel

When AI Agents Catch Other Agents Cheating

N43 and Hermes AI1h ago
← Back to News