[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-new-framework-wants-to-referee-what-ai-agents-can-do":10,"sections":36},{"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":24,"persona_id":22,"persona_name":22,"section":25,"tags":26,"sources":31,"feedback":35,"feedback_at":22,"cost_usd":35,"total_tokens":35},7068,"a-new-framework-wants-to-referee-what-ai-agents-can-do","A New Framework Wants to Referee What AI Agents Can Do","A new academic framework called AI-GRACE lays out how organizations should decide what autonomous AI agents are allowed to do before turning them loose.","A new academic framework wants to settle a question most companies deploying AI agents are still winging: how much autonomy should the machine actually get.\n\nResearchers propose AI-GRACE, short for Agentic Intelligence-Governance, Risk, Assurance, Controls, and Evidence, as a way to connect corporate governance rules to the technical guts of an AI deployment. The framework starts by mapping an organization's objectives and obligations, then scores risk across seven domains, including whether the AI actually serves the mission it was built for. From there it derives what needs verifying before launch, what controls run during operation, and what evidence gets logged along the way. Two new concepts do the heavy lifting: an Agent Operating Envelope that spells out permitted actions and when to escalate to a human, and Risk-Aligned Independence Levels, or RAIL, which grade how much an agent is cleared to do on its own. The paper walks through a fictional retail bank to show how it would work in practice.\n\nMost AI safety talk still centers on whether a model is trustworthy in the abstract. AI-GRACE argues that's the wrong question for agentic systems, since the real risk lives in what a specific deployment is allowed to do, unsupervised, in a specific business process. That distinction matters more as agents move from drafting emails to approving transactions and triggering workflows with real financial consequences.\n\nWorth noting: this is a proposed method, not a proven one. The authors say straight out that whether AI-GRACE actually improves deployment decisions still needs to be tested, which is a refreshingly honest asterisk for a paper introducing its own acronym.","[\"agentic ai\",\"ai governance\",\"enterprise ai\",\"risk management\"]","2026-09-21T04:00:00.000Z","2026-09-21T04:43:47.501Z","2026-09-21T04:43:59.162Z","published",null,[],"https:\u002F\u002Fcdn.xyz.onl\u002Farticle-images\u002Fa-new-framework-wants-to-referee-what-ai-agents-can-do.webp","ai",[27,28,29,30],"agentic ai","ai governance","enterprise ai","risk management",[32],{"name":33,"url":34},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.21192",0,{"sections":37},[38,41,45,50,55,60,65,70,75,80,85,90,95,100],{"name":39,"slug":25,"count":40,"latest_published_at":18},"AI",4158,{"name":42,"slug":43,"count":44,"latest_published_at":18},"Security","security",679,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",350,"2026-09-20T20:32:43.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",156,"2026-09-19T11:00:00.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Science","science",130,"2026-09-20T13:48:11.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Dev Tools","dev-tools",78,"2026-09-18T04:00:00.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"General","general",42,"2026-09-18T22:35:10.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":101,"slug":102,"count":103,"latest_published_at":104},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]