{
 "slug": "ai-agent-act-senator-warners-draft-legislation",
 "topic_id": "TOPIC-024",
 "cluster": "Agentic Commerce & MCP Tool-Call Gating",
 "tier": "Tier B",
 "title": "Reading draft agent legislation for architectural requirements",
 "summary": "Draft bills on autonomous software converge on a small set of demands. Most are architectural, most survive redrafting, and most are cheaper to build now than to retrofit.",
 "lede": "Legislative text changes between draft and enactment. What tends not to change is the underlying question a bill is trying to answer — and for agent legislation that question is consistently about accountability for automated acts.",
 "date": "2024-05-13",
 "category": "Future of Work",
 "author_id": "tobias-lindqvist-rao",
 "tags": [
  "AI legislation",
  "policy",
  "agent regulation",
  "compliance architecture",
  "accountability",
  "governance"
 ],
 "image_title": "Reading Draft Agent Legislation",
 "schema": "Article",
 "key_takeaways": [
  "Draft agent bills converge on identification, human accountability, records and revocation — across jurisdictions and drafting styles.",
  "Those four are architectural. Retrofitting them into a deployed agent estate is substantially harder than designing for them.",
  "Building to the convergent requirements is defensible engineering even if no specific bill passes, because the questions they encode get asked by counterparties and insurers regardless."
 ],
 "body": [
  {
   "type": "h2",
   "text": "Reading bills for the question, not the text"
  },
  {
   "type": "diagram",
   "kind": "chain",
   "alt": "Four convergent requirements, by retrofit cost",
   "caption": "The fourth cannot be created after the fact, because the human was never asked.",
   "nodes": [
    {
     "label": "Identification",
     "sub": "distinct agent identities",
     "note": "cheap",
     "good": true
    },
    {
     "label": "Records",
     "sub": "what executed",
     "note": "moderate",
     "good": true
    },
    {
     "label": "Revocation",
     "sub": "enforced, fails closed",
     "note": "moderate",
     "good": true
    },
    {
     "label": "Human accountability",
     "sub": "who authorised",
     "note": "cannot be backdated",
     "bad": true
    }
   ]
  },
  {
   "type": "p",
   "html": "Specific thresholds, definitions and penalties change repeatedly during drafting. The problem a bill is responding to changes rarely, and it is the problem that determines what you have to build."
  },
  {
   "type": "p",
   "html": "For agent legislation the problem is consistent: automated systems take consequential actions and nobody is clearly accountable for them."
  },
  {
   "type": "h2",
   "text": "The four convergent requirements"
  },
  {
   "type": "table",
   "head": [
    "Requirement",
    "What a bill typically asks",
    "What it means architecturally"
   ],
   "rows": [
    [
     "Identification",
     "Agents must be identifiable as such to counterparties",
     "A stable agent identity, distinguishable from a human's"
    ],
    [
     "Human accountability",
     "A named person or entity is responsible for the agent's acts",
     "A record connecting the agent to a principal, per action or per scope"
    ],
    [
     "Records",
     "Sufficient records to reconstruct an automated decision",
     "Execution records independent of the agent's own narration"
    ],
    [
     "Revocation",
     "Authority can be withdrawn and the withdrawal is effective",
     "Enforced revocation at the point of effect, not advisory"
    ]
   ]
  },
  {
   "type": "p",
   "html": "These appear in draft text from several jurisdictions with different legal traditions and different drafting conventions. The convergence is a signal that they are responses to the structure of the problem rather than one legislature's preference."
  },
  {
   "type": "h2",
   "text": "Which are cheap and which are not"
  },
  {
   "type": "ol",
   "items": [
    "<strong style=\"font-weight:600\">Identification is cheap.</strong> Issue distinct identities, label agent traffic, do not let agents share human accounts. Mostly hygiene you should want anyway.",
    "<strong style=\"font-weight:600\">Records are moderate.</strong> Capturing what executed, in canonical form, at the point of effect. Storage is trivial; the work is instrumenting the right points.",
    "<strong style=\"font-weight:600\">Revocation is moderate.</strong> Enforced at the effect rather than the gateway, failing closed. A day's work in a well-structured service, a month in a sprawling one.",
    "<strong style=\"font-weight:600\">Human accountability is expensive to retrofit.</strong> It requires a record connecting a named human to an action or a scope, produced at the time and verifiable afterwards. Systems that did not capture it cannot reconstruct it."
   ]
  },
  {
   "type": "p",
   "html": "The asymmetry in item four is the whole argument for acting before enactment. The other three can be added to a running system. The fourth cannot be backdated, because the human was not asked."
  },
  {
   "type": "h2",
   "text": "What \"human accountability\" does not mean"
  },
  {
   "type": "p",
   "html": "It does not mean a human approves every action. No draft requires that, and any that did would be unworkable."
  },
  {
   "type": "p",
   "html": "It means the chain from an action to a responsible human is traceable. For most actions that runs through a scope grant — a human authorised this class of action within these bounds. For a small set of high-consequence actions it runs through a per-action approval."
  },
  {
   "type": "p",
   "html": "Conflating the two produces either paralysis or a control that does not hold."
  },
  {
   "type": "h2",
   "text": "Building without betting on a bill"
  },
  {
   "type": "p",
   "html": "The honest framing: do not build to a draft. Build to the convergent requirements, which are demanded by parties other than legislatures."
  },
  {
   "type": "ul",
   "items": [
    "Counterparties ask whether your agent had authority before they perform",
    "Insurers ask what controls governed automated actions when pricing",
    "Auditors ask how automated decisions are evidenced under existing frameworks",
    "Your own incident response asks what actually executed"
   ]
  },
  {
   "type": "p",
   "html": "Each of those arrives regardless of whether any particular bill becomes law, and each is answered by the same four capabilities."
  },
  {
   "type": "h2",
   "text": "A reasonable sequence"
  },
  {
   "type": "ol",
   "items": [
    "Inventory agents and the consequential actions they can take. This is usually the step that surprises people.",
    "Give agents distinct identities. Stop credential sharing with human accounts.",
    "Instrument execution records at the point of effect for the consequential set.",
    "Implement enforced revocation with a short kill time.",
    "Add per-action human approval on the irreversible subset, bound to the rendered effect.",
    "Add scope grants for everything else."
   ]
  },
  {
   "type": "p",
   "html": "Steps one through four are defensible under any regulatory outcome. Steps five and six are where the effort concentrates, and where a system built without them will struggle later."
  },
  {
   "type": "p",
   "html": "This describes legislative tendencies at a general level and is not legal advice. Specific obligations depend on jurisdiction, sector and enacted text."
  },
  {
   "type": "h2",
   "text": "Why build without betting on a bill"
  },
  {
   "type": "table",
   "caption": "Who asks the same four questions regardless",
   "head": [
    "Party",
    "When"
   ],
   "rows": [
    [
     "Counterparties",
     "Before performing on an agent commitment"
    ],
    [
     "Insurers",
     "When pricing, and at claim"
    ],
    [
     "Auditors",
     "Under existing frameworks, now"
    ],
    [
     "<strong style=\"font-weight:600\">Your own incident response</strong>",
     "<strong style=\"font-weight:600\">At 3am</strong>"
    ]
   ]
  },
  {
   "type": "h2",
   "text": "Objections and honest limits"
  },
  {
   "type": "p",
   "html": "<strong style=\"font-weight:600\">“Human accountability means approving everything.”</strong> No draft requires that and it would be unworkable. It means the chain from an action to a responsible human is traceable — usually through a scope grant, per-action only on irreversible effects."
  },
  {
   "type": "p",
   "html": "<strong style=\"font-weight:600\">“We should wait for enacted text.”</strong> Identification, records and revocation can be added to a running system. Accountability cannot be backdated, which is the asymmetry that decides the sequencing."
  },
  {
   "type": "p",
   "html": "This describes legislative tendencies at a general level and is not legal advice. Obligations depend on jurisdiction, sector and enacted text."
  }
 ],
 "faq": [
  {
   "q": "Should we build to a bill that has not passed?",
   "a": "No. Build to the requirements that recur across drafts, because counterparties, insurers and auditors ask the same questions regardless of enactment."
  },
  {
   "q": "Does human accountability mean approving every action?",
   "a": "No. It means the chain from an action to a responsible human is traceable — usually through a scope grant, and through per-action approval only on irreversible effects."
  },
  {
   "q": "Which requirement is hardest to retrofit?",
   "a": "Human accountability. Identification, records and revocation can be added to a running system; a record of what a human authorised cannot be created after the fact."
  },
  {
   "q": "Is this legal advice?",
   "a": "No. It describes tendencies in draft legislation at a general level. Specific obligations depend on jurisdiction, sector and enacted text."
  },
  {
   "q": "Should we build to a draft bill?",
   "a": "No. Build to the requirements that recur across drafts, because counterparties, insurers and auditors ask the same questions regardless."
  },
  {
   "q": "Does accountability mean approving every action?",
   "a": "No. It means the chain from an action to a responsible human is traceable, usually via a scope grant."
  }
 ],
 "sources": [
  {
   "t": "Published draft legislation on autonomous software systems and AI agents."
  },
  {
   "t": "Regulation (EU) 2024/1689 (AI Act) — consolidated text",
   "u": "https://eur-lex.europa.eu/eli/reg/2024/1689/oj"
  },
  {
   "t": "Sector regulator guidance on accountability for automated decisions."
  },
  {
   "t": "Comparative analyses of AI governance requirements across jurisdictions."
  }
 ],
 "related": [
  {
   "slug": "my-agent-did-defense-businesses-will-dispute",
   "title": "The 'my agent did it' defence",
   "category": "Future of Work"
  },
  {
   "slug": "building-agent-kill-switch-actually-works-cryptographic-nonce",
   "title": "Building an agent kill-switch",
   "category": "Comparison"
  },
  {
   "slug": "fido-alliance-agentic-authentication-inside-race-standardize-verifiabl",
   "title": "Standardising verifiable user instructions",
   "category": "Future of Work"
  },
  {
   "slug": "section-5-ftc-act-autonomous-ai-agents-avoiding",
   "title": "Consumer protection liability when the bot makes the promise",
   "category": "Future of Work"
  }
 ],
 "image": "https://cdn.twc.sh/images/igcache/Reading%20Draft%20Agent%20Legislation/1500_900/blog.jpg",
 "wordcount": 939,
 "url": "/blog/ai-agent-act-senator-warners-draft-legislation.html",
 "reading_time": "4 min read",
 "meta_description": "Draft bills on autonomous software converge on a few demands. Most are architectural, survive redrafting, and are cheaper to build now.",
 "hub": {
  "slug": "topics/agent-tool-call-gating",
  "title": "Agent tool-call gating"
 },
 "answer": "The four demands that recur across every draft: identification, human accountability, records and revocation. Thresholds and definitions change repeatedly during drafting; the problem being responded to does not. Three of the four are cheap, and one cannot be retrofitted.",
 "answer_q": "What should you build from draft agent legislation?",
 "glossary": [
  {
   "term": "Convergent requirement",
   "def": "A demand appearing across drafts from different jurisdictions, indicating it responds to the problem's structure."
  },
  {
   "term": "Scope grant",
   "def": "A signed delegation covering a class of actions with bounds and an expiry."
  },
  {
   "term": "Retrofit cost",
   "def": "How expensive a capability is to add later — the variable that should drive sequencing."
  }
 ],
 "checklist": {
  "title": "A sequence that survives any outcome",
  "id": "sequence",
  "desc": "Six steps.",
  "steps": [
   {
    "name": "Inventory agents and their consequential actions.",
    "text": "The step most often skipped."
   },
   {
    "name": "Give agents distinct identities.",
    "text": "Stop credential sharing with human accounts."
   },
   {
    "name": "Instrument execution records at the point of effect.",
    "text": "For the consequential set."
   },
   {
    "name": "Implement enforced revocation.",
    "text": "Short kill time, fails closed."
   },
   {
    "name": "Add per-action approval on the irreversible subset.",
    "text": "Where effort concentrates."
   },
   {
    "name": "Add scope grants for everything else.",
    "text": "So the chain terminates at a person."
   }
  ]
 },
 "cta": {
  "title": "Where this fits in Manav",
  "html": "Manav covers three of the four directly: distinct agent identities under a signed human delegation, execution records at the point of effect, and revocation enforced where the effect happens.",
  "href": "../docs.html",
  "label": "See the primitives"
 }
}