Skip to content

GRF · AI and Digital Governance

Autonomous agents and liability

AI agents now plan and take actions across software, payments and physical systems with limited human input. When an agent causes harm, existing liability rules must decide who pays: the developer, the deployer, the user or the operator of the platform it acted on. This page sets out how autonomy strains liability and oversight, the records that show it, and who decides how responsibility is assigned.

01नोंदणी करा

Autonomous agents and liability on the record.

02The question

Autonomous agents and liability: how the risk develops.

Action boundaries and accountability. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.

यंत्रणा

Actions outrun intent

An agent pursuing a goal can take steps its user did not foresee or authorise. Fault-based liability struggles when no person intended or could reasonably foresee the harmful action.

यंत्रणा

Many hands, unclear fault

The model, the agent framework, tool providers, the deploying firm and the user all contribute. Proving which defect caused harm is costly for claimants without access to records.

यंत्रणा

Permissions define exposure

What an agent can access, spend or change sets the size of possible harm. Limits, confirmations and kill switches are governance choices made at deployment.

यंत्रणा

Law adapts slowly

Product liability, contract and tort law are being reinterpreted for software and AI. Until courts and legislatures settle questions, uncertainty itself shapes deployment.

03देखरेख करा

Public algorithm register.

The screen a team would use to follow this subject: its measures, states and decisions. Names are examples and values are illustrative.

Public algorithm register

Autonomous agents and liability · Algorithm register · Quarter 2

Two high-risk systems without assessmentGRF · AI and Digital Governance

Registered systems

Benefit fraud flaggingSocial agencyउच्चMissingPartial
Permit triagePlanningLimitedपूर्णहोय
School placementEducationउच्चपूर्णहोय
Chat assistantTax officeMinimalपूर्णलागू नाही

Oversight action

प्रणाली
Benefit fraud flagging
Required
impact assessment, appeal route
Deadline
77 days
प्राधिकरण
Oversight body

Complaints about automated decisions

Public algorithm register: the screen’s structure, terms and states for Autonomous agents and liability.Example names and illustrative values; a design, not a record or a live system.
04निर्देशक

What is measured, and how well.

Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.

IndicatorUnitCadenceSource typeLabel
Jurisdictions treating software as a product for liabilitycounton changeLegislationPublished statistics
Court decisions on AI-caused harmcount by typeannualCase law databasesPublished statistics
Agent incidents reportedcount by harm typemonthlyIncident databasesनोंदवलेले
Deployments with documented permission limitsshare of surveyed deploymentsperiodicSurveys and auditsDECLARED
Insurance products covering AI liabilitycount of offeringsannualMarket reportsनोंदवलेले
Regulatory guidance on agentic AIcount of documentsannualRegulator publicationsPublished statistics
05Instruments and commitments

How autonomous agents and liability is governed: instruments, mandates and commitments.

The instruments, mandates, actors and accountability arrangements a policymaker, diplomat or development agency works with, citing the agreements that set them. Structures follow the frameworks named. Any state or progress drawn is illustrative, not a record.

Policy instrument matrix

The kinds of instrument available on autonomous agents and liability, by level of government and cooperation.

Instrument typeLocalराष्ट्रीयप्रादेशिकजागतिक
RegulatoryZoning, building codesLaws, standards, licensingHarmonised rulesConventions and treaties
EconomicLocal levies, grantsBudget lines, taxes, insurance schemesRisk pools, regional fundsFinancing mechanisms
माहितीLocal warningsDisclosure, early warning, statisticsData sharing arrangementsGlobal monitoring and reporting
InstitutionalLocal committeesLead agencies, coordination councilsRegional bodies and secretariatsMultilateral mandates
CooperativeCommunity agreementsPartnership compactsMemoranda and joint programmesAgreements and joint commitments
A typology after the OECD Recommendation on the Governance of Critical Risks; examples of instrument types, not an inventory of any country’s measures.
Sendai Framework for Disaster Risk Reduction · UNDRRParis Agreement · UNFCCC2030 Agenda and the Sustainable Development Goals · United NationsInternational Health Regulations · WHO
06स्रोत

Where the evidence comes from.

Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.

स्रोतWhat it givesLicence class
EU revised Product Liability DirectiveLiability rules including software and AI systems as productsPublic, EU reuse terms
Law commission and parliamentary reports on AI liabilityAnalysis of liability options for AI and automated systemsPublic records, terms per body
OECD AI Incidents MonitorRecorded incidents involving AI systemsFree, OECD terms
UK Automated and Electric Vehicles Act and Automated Vehicles ActLiability allocation for automated vehiclesOpen Government Licence
Court judgment databasesDecisions on harm involving automated systemsPublic records, terms per court
07Decisions

Who decides, and what the evidence serves.

AI and Digital Governance prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.

Who decidesThe questionEvidence this page serves
Legislatures and law commissionsHow liability rules apply to autonomous systemsComparative regimes and case evidence
Deploying organisationsWhat an agent may do and what controls applyControl practice and incident evidence
InsurersWhether and how to cover AI liabilityIncident patterns and legal developments
CourtsWho is liable for harm in a specific caseRecords presented by parties; the page serves context only

Outputs this topic produces

  • Comparative notes on liability rules for software and AI
  • Guide to permission limits and controls for agent deployments
  • Dialogue briefs on allocating responsibility in multi-party AI chains
  • Questions for legislatures on agent liability
08Core questions

What experts ask about autonomous agents.

  • What actions is this agent permitted to take, and how are those limits enforced technically?
  • When an agent causes harm, who is liable: the developer, the deployer, the user?
  • How can an agent’s actions be traced, reviewed and reversed?
  • How should agents interact with other agents, and who governs those interactions?
09Mechanisms

Why agents raise new governance questions.

Most governance frameworks were built for systems that advise a person who then acts. Agents collapse that gap, and with it one of the main points at which errors were caught. Why now: AI agents that act on behalf of people and organisations raise questions of liability that existing rules did not foresee, and legislators are still working through them. The platform is designed to compare emerging approaches and to record open questions for decision holders, without giving legal advice.

Autonomy

Action without review

Agents can act faster and more often than people can review, so errors can multiply before anyone notices.

Interaction

Agents meeting agents

Interactions between automated systems can produce outcomes no one designed, as seen in automated trading.

Liability

Unclear responsibility

Existing liability rules assume a human or company acted; agents complicate who that is.

10Methods and limits

Instruments for bounding agents.

  1. 01

    Permission and scope controls

    Technical limits on what agents can access and do, studied under AI evaluation and assurance engineering. Limits: agents may find unintended routes around controls.

  2. 02

    Logging and reversibility

    Recording actions and designing them to be undoable where possible. Limits: some actions cannot be reversed.

  3. 03

    Human approval thresholds

    Requiring human sign-off above defined risk levels. Limits: can slow operations and invite approval fatigue.

  4. 04

    Liability analysis

    Comparing how legal systems allocate responsibility. Limits: law is still developing.

12प्रश्न

प्रश्न आणि उत्तरे.

What is an AI agent?
An AI system that pursues goals by planning and taking actions, such as calling tools, browsing, writing code or making payments, with limited step-by-step human direction.
Who is liable when an AI agent causes harm?
It depends on the jurisdiction and facts. Product liability, contract, negligence and sector rules may apply to developers, deployers or users; many questions are unsettled.
What controls limit agent harm?
Restricted permissions, spending limits, human confirmation for consequential actions, logging, monitoring and the ability to stop the agent.
Can an AI agent be legally responsible?
Under current law in most jurisdictions, no. Responsibility rests with people and organisations. How it is divided among them is still being worked out.
How can agent actions be kept within bounds?
Through technical permission limits, monitoring, human approval for high-risk actions and the ability to stop and reverse actions.
Does the platform test agents?
No. Technical evaluation is done at GCRI; the platform compares governance.
Why do agent interactions matter?
Because when many automated systems react to each other, small errors can amplify quickly, as sudden market moves driven by automated trading have shown.
अहवालयोगदान द्या

या पानावरून पुरावा योगदान द्या.

Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.

Global Risks ForumAI and Digital GovernanceExponential TechnologyEvidence Brief

वैयक्तिक डेटा, संवेदनशील पायाभूत सुविधांचा तपशील किंवा सामायिक करण्याचा तुम्हाला अधिकार नसलेले काहीही कधीही पाठवू नका. योगदानाचे वापरापूर्वी पुनरावलोकन केले जाते; ते पाठवल्याने कोणतेही लेखकत्व किंवा संपादकीय अधिकार मिळत नाहीत, आणि प्रायोजकत्वाने कधीही निष्कर्ष विकत घेता येत नाही.

AI आणि डिजिटल प्रशासन

Bring a governance question to the Forum.

Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.

कायदेशीर
Show all 18 legal documents

या साइटवरील नकाशे, डेटा आणि आकृत्या संदर्भ आणि सर्वसाधारण माहितीसाठी दाखवल्या आहेत. त्या सल्ला, अंदाज किंवा इशारे नाहीत; माहितीच्या मर्यादा अस्वीकरणे आणि डेटा सूचनांमध्ये नमूद केल्या आहेत.