[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-researchers-build-an-audit-test-for-ai-discovery-claims":10,"sections":34},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},6316,"researchers-build-an-audit-test-for-ai-discovery-claims","Researchers Build an Audit Test for AI Discovery Claims","A new certification protocol forces AI research agents to prove their results can't be trivially recovered before anyone calls them a discovery.","A new audit framework wants proof before anyone calls an AI research agent's output a genuine discovery.\n\nResearchers built the Discovery Certification Protocol (DCP), a set of executable tests for claims that an AI research agent found something genuinely new. The protocol runs matched agents through the same starting information and public web content the original agent had, but withholds its actual research history; any independent agent that still reaches the same result triggers a veto on the discovery claim. A separate optional test compares how much truthful feedback helps an agent versus a neutral baseline policy. The team ran two audits, one on SQLite query optimization and one on a simulated catalyst-design task; each produced zero recoveries across 96 episodes with a statistical upper bound of 0.0468, while feedback tests logged 30 successful recoveries with truthful signals against zero with neutral ones.\n\nAI labs and startups routinely tout agents that discover optimizations, molecules, or proofs, but a high benchmark score does not rule out an agent stumbling onto something already implicit in its training data or the public web. DCP gives outside auditors a reproducible, deterministic way to check that claim instead of trusting a leaderboard number. It is the kind of adversarial preregistration that fields like clinical trials adopted decades ago, and AI research has mostly lacked.\n\nTwo case studies and 96 episodes is a small sample to hang a new evidentiary standard on, and the real test will be whether any AI lab lets outsiders run this audit on a claim it actually cares about.","[\"ai\",\"ai-research\",\"reproducibility\",\"benchmarks\"]","2026-09-11T04:00:00.000Z","2026-09-11T06:01:56.275Z","2026-09-11T06:02:08.168Z","published",null,[],"ai",[24,26,27,28],"ai-research","reproducibility","benchmarks",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.09219",0,{"sections":35},[36,39,43,47,52,57,62,65,70,74,79,84,89,94],{"name":37,"slug":24,"count":38,"latest_published_at":18},"AI",3522,{"name":40,"slug":41,"count":42,"latest_published_at":18},"Security","security",637,{"name":44,"slug":45,"count":46,"latest_published_at":18},"Policy","policy",338,{"name":48,"slug":49,"count":50,"latest_published_at":51},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":53,"slug":54,"count":55,"latest_published_at":56},"Hardware","hardware",153,"2026-09-09T15:12:32.000Z",{"name":58,"slug":59,"count":60,"latest_published_at":61},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":63,"slug":64,"count":60,"latest_published_at":18},"Science","science",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":18},"Dev Tools","dev-tools",70,{"name":75,"slug":76,"count":77,"latest_published_at":78},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"General","general",41,"2026-09-08T01:57:23.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]