Draft:Machine Economy

Economic systems involving autonomous machine-to-machine interactions From Wikipedia, the free encyclopedia

Machine economy is a term used to describe economic systems in which machines, connected devices, software agents, artificial intelligence (AI) systems, robots, or cyber-physical systems participate in economic interactions with limited direct human intervention. Such interactions may include exchanging data, negotiating terms, triggering contracts, making or receiving payments, providing services, consuming services, and acting under delegated authority from a human or legal person.[1][2]

  • Comment: In accordance with Wikipedia's Conflict of interest guideline, I disclose that I have a conflict of interest regarding the subject of this article. Carsten.stoecker (talk) 12:15, 9 May 2026 (UTC)


The term is used in academic and technical literature on the Internet of things (IoT), artificial intelligence, blockchain, digital platforms, cyber-physical systems, multi-agent systems, and autonomous agents. Academic literature describes machine economies as systems involving economically autonomous machines, or as an emerging phenomenon enabled by the convergence of IoT, AI, blockchain, digital platforms, and machine-to-machine coordination.[1][3][4]

The machine economy is not a single technology stack. It is a broad concept at the intersection of automation, AI, IoT, digital identity, payment infrastructure, smart contracts, market design, law, taxation, cybersecurity, and governance. Some components already exist in controlled form, such as machine-to-machine communication, automated contracting, programmable payments, device certificates, smart charging, industrial automation, and transactive energy systems. More advanced forms, in which machines autonomously negotiate and transact across open markets, remain emerging or speculative.

Definition and scope

A machine economy can be defined as an economic environment in which machines, connected devices, software agents, AI systems, robots, or cyber-physical systems identify themselves, exchange data, make decisions, and participate in value exchange under defined rules and authority.[1][3]

The term is related to, but distinct from, automation. Automation can occur without economic exchange between machines. It is also broader than machine-to-machine payments, which concern only payment execution. A machine economy may include identity, authority, communication, contracting, settlement, auditability, liability, taxation, and governance.

Academic descriptions differ in emphasis. Some sources define the machine economy through economically autonomous machines. Others focus on the convergence of IoT, AI, blockchain, and digital platforms. Industry sources often stress connected devices acting as market participants. These definitions overlap, but they do not always agree on whether blockchain is necessary.[2][3][4]

Significance

The concept has gained relevance as connected devices, digital payments, AI agents, industrial automation, and cyber-physical infrastructure become more closely integrated. It provides a common term for economic interactions in which non-human technical systems perform actions that previously required direct human or organizational intervention.

Research on machine economies also connects technical questions with economic and legal ones. Hartwich et al. describe efficient machine economies in relation to autonomous exchange of information and value, transaction costs, property rights, governance, and Pareto efficiency.[1] Jöhnk et al. describe the machine economy as a phenomenon requiring conceptual understanding of driving technologies and their interrelations.[2] Duda et al. describe it as an emerging phenomenon combining IoT, AI, and blockchain to enable economically autonomous acting machines.[3]

The term is relevant for several current policy and technology discussions, including machine-to-machine payments, delegated AI agents, autonomous vehicles, smart charging, transactive energy, trusted digital identity, secure wallets, post-quantum migration, product liability, and automated contracting.

Terminology and related concepts

The machine economy overlaps with several established concepts.

The Internet of things provides connected sensors, devices, and embedded systems. IoT can support the physical and communication layer for machine transactions. Cyber-physical systems combine computation, networking, and physical processes. They are important where machine decisions affect vehicles, factories, energy systems, robots, medical devices, or other physical infrastructure.

Software agents and multi-agent systems can act in digital environments by calling APIs, retrieving data, triggering workflows, or interacting with other agents. In newer AI systems, agents may use planning, tool use, and multi-step execution.

Smart contracts are programs that execute transaction logic. They may support machine transactions, but they do not by themselves solve identity, authority, safety, legal attribution, taxation, or liability. Machine-to-machine payments are a subset of the machine economy, not a synonym for it.

Related terms include machine-to-machine economy, M2M economy, machine-to-everything economy, economy of things, IoT economy, agent-based economy, and cyber-physical economy.[5]

History and early uses

The expression "machine economy" has older meanings that predate its contemporary use in relation to AI, IoT, blockchain, and autonomous machine-to-machine transactions. In nineteenth- and early twentieth-century sources, it was mainly used to describe mechanised production, factory organisation, large-scale industry, and the social effects of the machine age.

A conceptual antecedent is Charles Babbage's 1832 book On the Economy of Machinery and Manufactures, which analysed the economic organisation of machinery and manufacturing, although it did not use the contemporary meaning of autonomous machine transactions.[6]

One verified late-nineteenth-century use appears in the work of J. A. Hobson. Hobson used the hyphenated form "machine-economy" in an 1896 article on collectivism in industry.[7] Hobson also used the phrase in early editions of The Evolution of Modern Capitalism: A Study of Machine Production, originally published in 1894 and later revised.[8] In Imperialism: A Study (1902), Hobson used the term in relation to advanced industrial methods and the expansion of markets.[9]

A Depression-era cluster of usage appeared around 1932, during public debates on technological unemployment, the "machine age", industrial abundance, and Technocracy. Glenn Frank's Thunder and Dawn used the expression in this context; a contemporary review described chapters in which "our machine economy" was defended as a possible tool of emancipation.[10] Historical accounts of the Technocracy movement describe 1932–1933 as the period in which machine-age planning, unemployment, and industrial productivity became especially prominent in American public debate.[11] Lewis Mumford's Technics and Civilization also used the phrase in a broader critique of technology and industrial society.[12]

Postwar social criticism continued to use the term in the older industrial sense. Paul and Percival Goodman used it in Communitas, and Paul Goodman used it in Growing Up Absurd in discussion of social organisation under modern industrial production.[13][14]

The contemporary meaning developed later from several research and technology streams: agent-based computational economics, computational markets, market-based control, IoT, cyber-physical systems, smart grids, blockchain, and AI agents. In the 1990s, Michael P. Wellman described market-oriented programming and computational markets, showing how software agents could use price systems for distributed resource allocation.[15][16]

By the early 2020s, information-systems literature treated the machine economy as a research topic in its own right.[2][1][3]

Technical foundations

Machine economy systems may include several technical layers.

At the physical layer, connected machines require sensors, actuators, embedded systems, network interfaces, and secure firmware. At the autonomy layer, they may use rule-based control, optimization, AI agents, or multi-agent systems. At the communication layer, they require APIs, industrial protocols, IoT protocols, or data-space connectors.

At the trust layer, machines require identifiers, credentials, keys, authorization policies, audit logs, and revocation mechanisms. At the transaction layer, they may use conventional payments, instant payments, tokenized deposits, stablecoins, payment channels, smart contracts, or other settlement systems. At the governance layer, rules define who may act, under what authority, with what limits, and with what liability.

Not all machine economy systems require blockchain. Blockchain can be useful for shared records, programmable settlement, distributed coordination, or tamper-evident state. Many industrial and regulated systems may instead use conventional databases, APIs, regulated payment providers, data-space governance, public key infrastructure, or sector-specific trust frameworks.

Types and governance

Hartwich et al. distinguish machine economies by two dimensions: whether interactions are machine-to-human or machine-to-machine, and whether governance is enforced by humans or machines.[1] This creates four broad types:

  • machine-to-human economies under human governance;
  • machine-to-machine economies under human governance;
  • machine-to-human economies under machine governance; and
  • machine-to-machine economies under machine governance.

The distinction is important because governance affects economic efficiency, accountability, dispute resolution, and risk control. Human governance may rely on contracts, courts, regulators, firms, platforms, or human operators. Machine governance may rely on software rules, smart contracts, automated policy engines, credentials, or protocol-level enforcement.

Machine governance remains limited. Smart contracts and automated rules can enforce some technical conditions, but they do not by themselves solve questions of legal responsibility, safety, fairness, taxation, revocation, or dispute handling.

AI agents and software autonomy

Software agents can act in digital environments by calling APIs, retrieving data, triggering workflows, buying digital services, or negotiating with other systems. In a machine economy, a software agent may act for a person, company, platform, robot, vehicle, or industrial asset.

Newer AI systems increase this relevance because some agents can plan, use tools, and execute multi-step actions. This also creates risks. Prompt injection, tool misuse, unauthorized delegation, and weak access controls can allow an agent to act outside its intended scope.[17]

In law and governance, a software agent is usually not treated as an independent legal person. It is normally a system operated by, or attributable to, a human or legal entity. Therefore, machine economy systems need clear authorization, spending limits, logs, and revocation mechanisms.

Physical AI and autonomous machines

Physical AI systems differ from software-only agents because their actions may affect the physical world. A digital agent can cause financial, privacy, or cybersecurity harm. A robot, autonomous vehicle, drone, battery, medical device, or industrial machine may also cause bodily injury, property damage, grid instability, environmental harm, or production failures.

A 2026 Bitkom report describes Physical AI as AI systems that process information from physical environments, make decisions, and act in real-world processes through the interaction of AI, sensors, robotics, and software-supported decision processes.[18] Physical AI is relevant to machine economy discussions because machines that act in the physical world may also request services, purchase energy, sell data, pay fees, or coordinate with other autonomous systems.

Possible machine economy examples include electric vehicles buying charging services, batteries selling flexibility services, smart meters supporting energy-market participation, industrial robots requesting maintenance, drones paying for landing or charging, and autonomous logistics systems paying tolls or service fees.

Electric vehicle charging is one of the clearest bounded examples. ISO 15118 Plug & Charge allows automatic authentication and contract-based charging processes between an electric vehicle and charging infrastructure. The Open Charge Alliance has described how ISO 15118 certificate management and Plug & Charge can be combined with OCPP-based charging infrastructure.[19]

Identity, delegation, and trust infrastructure

Digital identity is a core issue in machine economy systems. A system must determine which machine or agent is acting, who controls it, what it is allowed to do, which organization is responsible, which credentials it holds, and whether those credentials remain valid.

Machine identity may include device identity, robot identity, service identity, API identity, vehicle identity, battery identity, smart meter identity, wallet identity, and digital twin identity. W3C decentralized identifiers are relevant because DIDs can refer to people, organizations, things, data models, or abstract entities.[20]

Verifiable credentials are also relevant. The W3C Verifiable Credentials Data Model defines a way to express tamper-evident claims made by an issuer and exchanged through an issuer-holder-verifier model.[21] Several protocol families are used or proposed for credential issuance, presentation, and exchange. OpenID for Verifiable Credential Issuance defines an API for issuing verifiable credentials, and OpenID for Verifiable Presentations defines mechanisms for requesting and presenting credentials.[22][23]

DIDComm Messaging is another protocol family used in decentralized identity systems. The DIDComm Messaging v2.0 specification describes a secure and private communication method built on DIDs.[24] DIDComm-based protocols include Issue Credential 2.0, which formalizes messages used to issue credentials, and Present Proof 2.0, which supports verifiable presentation exchange.[25][26] Hyperledger Aries also defines related issue-credential and present-proof protocols for exchanging verifiable credentials and presentations over DIDComm.[27][28]

Machine identity alone is not enough. In most practical systems, a machine must be linked to a legal person, owner, operator, manufacturer, deployer, or service provider. It must also prove delegated authority. For example, a machine may need to prove not only that it is a certified device, but also that it is authorized by a specific company to buy a service, sell data, sign a usage event, or trigger a payment.

European machine-identity and device-identity initiatives

Industrial data-space projects illustrate parts of this model. Gaia-X uses verifiable credentials and signed descriptions as part of its trust framework.[29] Catena-X describes identity wallets that hold proof of identity, credentials, roles, rights, services, and legal entity information.[30]

Other European data-space initiatives apply similar concepts in sector-specific contexts. Energy data-X aims to create a Gaia-X-based data space for the German energy sector and for standardized, sovereign data exchange among energy-market participants.[31] Manufacturing-X is a German and European initiative for trusted data ecosystems in manufacturing based on open standards.[32] Chem-X aims to establish an interoperable data ecosystem for the chemical industry, including technical standards for secure and efficient data exchange.[33] Aerospace-X is a Manufacturing-X project for aerospace supply chains that aims to develop a federated data-space platform using modern data-space technologies.[34] These initiatives are not machine economies by themselves, but they show how sectoral data infrastructures use identity, credentials, policies, and governance mechanisms that may also be relevant for machine-to-machine transactions.

In Europe, the European Commission has proposed a regulation on European Business Wallets, intended to support secure and trusted digital identification for economic operators across borders.[35] Such infrastructure may become relevant where companies delegate authority to software agents, machines, or digital services.

Non-European machine-identity and device-identity initiatives

Several non-European initiatives address machine identity, device identity, industrial identifiers, or trusted onboarding of connected devices. These projects are relevant to machine economy systems because autonomous machine-to-machine transactions require reliable identification, authentication, authorization, and lifecycle management.

In the United States, the National Cybersecurity Center of Excellence at the National Institute of Standards and Technology has developed guidance on Trusted IoT Device Network-Layer Onboarding and Lifecycle Management. The project focuses on providing network credentials to IoT devices in a trusted manner and maintaining secure device posture throughout the device lifecycle.[36][37]

In China, the Industrial Internet Identifier Resolution System has been developed as core infrastructure for the industrial internet. It is intended to identify and resolve industrial entities such as equipment, products, components, and digital resources, and has been promoted by Chinese industrial-internet policy and implementation bodies.[38][39] Analyses of China's IoT strategy have described the system as an industrial identifier infrastructure operating in parallel with the global internet's addressing system.[40]

In Japan, the Trusted Web initiative and the Ouranos Ecosystem address trust, data verification, and industrial data sharing. The Japanese Digital Agency describes Trusted Web as an initiative to strengthen data control and enable verification of exchanged data and counterparties.[41] The Ministry of Economy, Trade and Industry describes the Ouranos Ecosystem as an initiative for service-driven data spaces, and related architecture documents include identity, authentication, authorization, and trust functions.[42][43]

In South Korea, oneM2M-based IoT platforms and related open-source implementations have been used to support interoperable machine-to-machine and IoT services. The Mobius platform, developed by the Korea Electronics Technology Institute, is an open-source IoT server platform based on oneM2M that provides common service functions such as registration, data management, subscription and notification, and security.[44] Academic work on device identification interoperability in heterogeneous IoT platforms has used Mobius and related platforms to discuss identification, authentication, and authorization of devices and users across IoT systems.[45]

Emerging AI and physical AI identity initiatives

Emerging AI-agent and Physical AI identity initiatives are developing in Europe and outside Europe. In Europe, the WE BUILD consortium has published a non-paper on trusted identities for AI agents, digital wallets, and payments, arguing that AI agents need verifiable trust chains when acting in digital transactions.[46] The European Commission's European Business Wallet proposal is also relevant because it aims to provide trusted digital identification for economic operators and could support delegated actions by companies, software agents, machines, or digital services.[35] In Physical AI, Bitkom describes systems that perceive, understand, reason about, and act in the physical world, while the World Economic Forum describes Physical AI as a new stage of industrial operations based on advances in robotics, sensors, hardware, and AI.[18][47] Outside Europe, related work includes the United States National Institute of Standards and Technology project on trusted IoT device onboarding and lifecycle management, Japan's Trusted Web and Ouranos Ecosystem work on trusted data exchange and data spaces, China's Industrial Internet Identifier Resolution System, and South Korea's oneM2M-based IoT platform work. These initiatives do not constitute machine economies by themselves, but they show growing attention to identity, onboarding, authorization, verification, and lifecycle management for AI agents, connected devices, and cyber-physical systems.[36][41][42][39][45]

Machine-to-machine payments are often presented as a key use case. The Deutsche Bundesbank lists machine-to-machine payments, IoT payments, and pay-per-use payments as examples of programmable payment applications. It gives examples such as an electric car paying a charging station or parking facility, and a leased machine billing usage units independently.[48]

Payment options may include conventional bank payments, instant payments, card payments, stablecoins, tokenized deposits, commercial bank money tokens, central bank digital currencies, payment channels, smart-contract escrow, or conditional settlement. Bank for International Settlements work on tokenization discusses tokenized deposits, central bank digital currencies, programmable platforms, and atomic settlement.[49]

Blockchain became associated with the machine economy because it can support programmable settlement, shared ledgers, smart contracts, wallets, and peer-to-peer coordination. Academic work often names blockchain alongside IoT and AI as one of the important technology fields in machine economy research.[2][4]

Blockchain systems can also be used as registries for machine, service, or agent discovery. In such designs, a smart-contract registry may store or reference identifiers, service endpoints, capabilities, cryptographic verification material, reputation records, or validation results. For example, ERC-1056 describes an Ethereum-based lightweight identity registry for key and attribute management, including delegates and attributes associated with an identity, and is intended to support DID-compatible identity management.[50] ERC-8004, a draft Ethereum standard for "trustless agents", proposes on-chain identity, reputation, and validation registries to support discovery of agents and the establishment of trust through reputation and validation. Its registration files may describe agent services, endpoints, and interaction methods, while validation records may reference off-chain evidence or audits of agent work.[51] These mechanisms can contribute to trust domains in which machines or agents establish identity and present verification, validation, or behavioural evidence. They do not by themselves prove that an advertised capability is safe, functional, compliant, or legally authorized.

However, blockchain is not sufficient by itself. Autonomous payments also require identity, authorization, secure wallets, compliance checks, spending limits, dispute handling, revocation, audit logs, taxation records, and liability rules.

Economic models and market design

The machine economy is linked to agent-based computational economics, computational markets, market-based control, transaction cost economics, platform economics, auction theory, mechanism design, and the economics of automation. These fields provide different ways to analyse how machines, software agents, AI systems, and cyber-physical systems may coordinate, exchange value, and allocate resources.

Agent-based economics

Agent-based computational economics studies economic processes as dynamic systems of interacting agents.[52] This is relevant because machine economy systems may include large numbers of autonomous or semi-autonomous agents that repeatedly interact, learn, bid, negotiate, and adapt.

Computational market research shows how software agents can coordinate through prices. Wellman's work on market-oriented programming described how market price systems can support distributed resource allocation among computational agents with limited communication overhead.[15][16] In machine economy systems, similar ideas may apply to resource allocation among machines, services, batteries, vehicles, sensors, robots, and software agents.

Microeconomic concepts

At the microeconomic level, machine economy systems raise questions about transaction costs, property rights, information asymmetry, incentives, agency, and market design. Hartwich et al. discuss efficient machine economies by drawing on the Coase theorem and translating economic criteria into technical requirements for systems in which non-human agents exchange information and value.[1] Coase's work on social cost and property rights is relevant because many machine transactions depend on clearly defined rights, low transaction costs, and enforceable agreements.[53]

Transaction cost economics is also relevant because machine transactions may reduce some costs of search, contracting, monitoring, and settlement, while creating new costs for identity, verification, cybersecurity, dispute resolution, and interoperability. Williamson's transaction cost approach explains how markets, firms, and hybrid governance forms can be compared by their contracting and governance costs.[54]

Information asymmetry can arise when a machine, service, dataset, AI model, or physical device claims a capability or quality that counterparties cannot directly verify. Akerlof's "market for lemons" model is relevant because low-quality or falsely described assets can reduce trust and market participation when quality is hard to observe.[55] Machine-verifiable credentials, provenance records, testing results, and certifications may reduce such information asymmetry, but they also introduce new governance and verification requirements.

Agency problems are also relevant. In most legal systems, a machine or AI agent acts on behalf of a human, firm, owner, operator, or platform rather than as an independent legal person. Principal-agent theory is therefore relevant to questions of delegated authority, incentive alignment, monitoring, and liability.[56] A machine may be technically able to transact, but economic design must define the principal, the agent's authority, the limits of action, and the allocation of gains and losses.

Market design and platforms

Machine economy systems may use auctions, dynamic pricing, bilateral negotiation, posted prices, subscription models, pay-per-use models, or automated procurement. These mechanisms can support the allocation of scarce resources such as energy, bandwidth, compute, storage, data, charging capacity, logistics capacity, or machine time.

Some machine economy systems may be organized through platforms rather than open peer-to-peer markets. Platform economics is relevant because platforms may coordinate multiple sides of a market, set access rules, define fees, provide identity and reputation systems, and control data flows. Rochet and Tirole's work on two-sided markets explains how platforms must bring different sides of a market on board and how pricing structures can differ across sides.[57]

Market-design questions include how machines discover counterparties, how capabilities are verified, how prices are formed, how disputes are resolved, how malicious agents are excluded, how collusion is detected, and how market power is limited. In autonomous machine-to-machine settings, these questions may need to be answered by protocols, credentials, policies, audits, and automated monitoring rather than by manual review alone.

Macroeconomic concepts

At the macroeconomic level, the machine economy is related to automation, capital deepening, productivity, labour substitution, data-driven coordination, and the distribution of income between labour, capital, and platform owners. These effects are not specific to machine-to-machine transactions, but machine economy systems may intensify them where autonomous machines perform tasks, allocate resources, or generate revenue with reduced human involvement.

Automation research has found that robots and other computer-assisted technologies can affect employment and wages. Acemoglu and Restrepo found negative effects of industrial robot exposure on employment and wages in U.S. local labour markets.[58] Earlier work by Autor, Levy, and Murnane linked computerization to changes in the task content of work, especially through the substitution of routine tasks.[59]

Machine economy systems may also affect productivity and resource utilization by reducing coordination costs and enabling more granular allocation of assets. Examples include automated use-based billing, dynamic energy pricing, machine-to-machine settlement, and automated allocation of compute, data, logistics, or charging capacity. These benefits depend on reliable identity, interoperability, cybersecurity, market liquidity, and governance.

Macroeconomic risks include concentration of market power, platform dependency, reduced bargaining power for labour, new forms of systemic cyber risk, and the concentration of machine-generated data and revenue in firms that control infrastructure, platforms, or models. The distribution of gains from machine autonomy may therefore depend on ownership, regulation, taxation, competition policy, and access to trusted infrastructure.

Energy and resource markets

Energy systems provide one practical field for machine economy concepts. NIST describes transactive energy as a system of economic and control mechanisms that balances supply and demand across electrical infrastructure using value as an operational parameter.[60] IRENA describes peer-to-peer electricity trading as an online marketplace in which prosumers and consumers can trade electricity.[61]

In such systems, machines may include smart meters, electric vehicles, batteries, heat pumps, solar inverters, grid assets, and energy management systems. Market design must address not only prices and incentives, but also grid stability, regulatory duties, data access, identity, cybersecurity, and operational safety.

Open economic questions

Important market-design questions include whether machines are economic actors or tools, who owns machine-generated data and revenue, who bears risk, how delegated authority is limited, how taxation applies, and how systems prevent collusion, fraud, market manipulation, or unsafe optimization. Other open questions include whether machine-to-machine markets should be governed by platforms, public infrastructure, regulated data spaces, smart contracts, or hybrid institutional models.

Legal status, liability, and taxation

Machines, robots, and AI agents are generally not treated as legal persons. In most legal systems, they are products, tools, assets, services, or software systems operated by legal persons. Legal personhood for AI and robots remains a debated academic topic rather than the default legal position.[62][63][64]

Automated contracting is already recognized in some legal frameworks. The United States E-SIGN Act provides that a contract or record may not be denied legal effect solely because its formation, creation, or delivery involved electronic agents, if the action is legally attributable to the person to be bound.[65] The Uniform Electronic Transactions Act provides that a contract may be formed by interaction of electronic agents even if no individual reviewed the actions or resulting terms.[66]

UNCITRAL adopted the Model Law on Automated Contracting in 2024 to support legal recognition of automation, AI techniques, smart contracts, and machine-to-machine transactions in contract formation and performance.[67]

In the European Union, the AI Act defines an AI system as a machine-based system designed to operate with varying levels of autonomy and to generate outputs that may influence physical or virtual environments.[68] The revised Product Liability Directive extends no-fault liability for defective products to all movables, including software, and treats cybersecurity vulnerabilities as relevant to defectiveness.[69]

Taxation

Taxation issues arise because machines and AI agents may generate revenue, trigger payments, consume services, create data, or allocate costs across borders. In current tax systems, income and tax obligations are generally attributed to legal persons, such as owners, operators, employers, contractors, or platform providers, rather than to machines themselves.

The taxation of automation and robots has been debated mainly in relation to labour displacement and the tax base. Abbott and Bogenschneider argue that tax systems may favour automation when labour and capital are treated differently, and they discuss robot-tax proposals as a possible response to automation.[70] Thuemmel's work on optimal taxation of robots finds that robot taxes or subsidies depend on general-equilibrium effects and that changes to labour-income taxation may deliver larger welfare gains than a separate robot tax.[71]

For machine economy systems, tax administration may require auditable transaction records, reliable identification of the legal persons behind machines, place-of-supply information for VAT and sales tax, and rules for attributing income, expenses, and risk. OECD work on the digital economy notes that digitalization creates challenges for international tax rules and that it is not feasible to ring-fence the digital economy from the rest of the economy for tax purposes.[72] OECD VAT/GST guidelines also address cross-border trade in services and intangibles, which is relevant where machines or agents buy, sell, or consume digital services across jurisdictions.[73]

Security, functional safety, and machine supply-chain transparency

Machine economy systems may concentrate economic authority in software agents, devices, cryptographic keys, credentials, wallets, APIs, and smart contracts. This creates risks such as machine identity theft, unauthorized delegation, wallet compromise, key theft, smart-contract bugs, payment fraud, false provenance, data poisoning, prompt injection, firmware compromise, remote-attestation failure, and revocation failure.

Cyber-physical risk

For physical systems, cybersecurity, endpoint security, and functional safety are closely linked. A compromised robot, vehicle, battery, smart meter, medical device, or industrial machine may create physical harm, not only data loss. NIST's AI Risk Management Framework is relevant because it is designed to help manage AI risks to individuals, organizations, and society.[74]

Functional safety standards address the safety-related behaviour of electrical, electronic, programmable electronic, and control systems. IEC 61508 is a basic functional safety standard for electrical, electronic, and programmable electronic safety-related systems; ISO 26262 applies functional-safety principles to road vehicles; and ISO 13849 specifies requirements and guidance for safety-related parts of machinery control systems.[75][76][77]

Supply-chain security

Supply-chain security is also relevant to machine economy systems. Machines and AI agents may depend on software components, AI models, datasets, firmware, sensors, hardware modules, cloud services, and physical devices from multiple suppliers. Weak provenance or compromised components can affect whether a machine's identity, behaviour, outputs, or transactions can be trusted.

NIST's cybersecurity supply-chain risk-management guidance describes practices for identifying, assessing, and managing cybersecurity risks across suppliers, products, and services.[78] NIST's Secure Software Development Framework addresses secure software development practices for reducing software vulnerabilities, while the Supply-chain Levels for Software Artifacts framework defines controls for software supply-chain integrity and provenance.[79][80]

Machine-verifiable passports

An emerging architectural concept is a machine-verifiable cyber-physical system entity passport. Such a passport could describe a device, machine, robot, vehicle, AI service, software agent, or digital twin and provide signed evidence about its identity, owner or operator, capabilities, software and firmware versions, AI model lineage, safety certifications, test results, maintenance history, data provenance, authorization status, and operational limits. Related examples include device passports, digital product passports, and proposed AI service passports. In a machine economy, such passports could support the onboarding of a new counterparty machine in a machine third-party risk-management process.

The same basic controls used in traditional economies, such as counterparty onboarding, supply-chain transparency, risk scoring, certification, compliance checks, and transaction monitoring, may also be needed in cyber-physical machine economies. The difference is the time scale and degree of automation. In traditional business processes, supplier onboarding, third-party risk review, certification checks, and contractual approval can take days, weeks, or months. In a machine economy, comparable checks may need to be performed automatically and close to real time, similar to straight-through processing, before a machine, device, or AI agent is allowed to transact.

Provenance graphs

Machine-verifiable passports and provenance graphs are possible mechanisms for automated onboarding. A provenance graph can link evidence about a machine, model, software component, dataset, physical device, maintenance event, certification, or operational behaviour. The W3C PROV Ontology provides a semantic model for representing provenance information, and the W3C Verifiable Credentials Data Model provides a model for tamper-evident claims that can be exchanged between issuers, holders, and verifiers.[81][21]

Identity wallets, including organizational or business wallets, could be used to hold and present such claims in contexts where a legal person, machine, service, or AI agent must prove authorization, provenance, certification, or compliance.[35]

Semantic interoperability

Semantic interoperability is important in such systems because independent machines, suppliers, operators, regulators, and service providers may need to interpret the same evidence across organizational and sector boundaries. Provenance standards such as W3C PROV and content-provenance systems such as C2PA illustrate how cryptographically bound assertions can be used to express origin, modification history, and authenticity for digital assets, although they do not by themselves determine legal liability or safety.[81][82]

Zero trust and endpoint security

Cybersecurity, endpoint security, and zero trust architecture are also relevant to machine economy systems. Machine-to-machine transactions may involve large numbers of distributed endpoints, including sensors, robots, vehicles, smart meters, gateways, edge devices, APIs, and AI services. Each endpoint can become an attack surface for credential theft, malware, lateral movement, data manipulation, or unauthorized transactions.

NIST Special Publication 800-207 defines zero trust architecture as an approach in which access decisions are based on continuous evaluation of identity, device posture, policy, and context, rather than implicit trust based on network location.[83] CISA's Zero Trust Maturity Model describes identity, devices, networks, applications and workloads, and data as core pillars for zero trust implementation.[84]

Applied to a machine economy, zero trust concepts imply that machines and agents should be continuously authenticated, authorized, monitored, and risk-scored before and during interactions, rather than trusted only because they are inside a network, factory, fleet, or data space.

Functional safety

Functional safety adds a separate but related requirement. Cybersecurity controls aim to prevent unauthorized access, manipulation, or disruption, while functional safety aims to ensure that safety functions work correctly or fail in a predictable and safe way. In cyber-physical machine economies, the two domains can interact: a cyberattack may defeat a safety function, and an unsafe autonomous decision may cause physical harm even without a malicious attack.

For high-risk systems, transaction authorization may therefore need to account not only for identity and payment risk, but also for the current safety state of the machine, the validity of safety certifications, operating conditions, and whether the requested action is within approved operational limits.

Risk controls

Risk controls may include hardware-backed keys, endpoint detection and response, secure boot, signed firmware, scoped credentials, least privilege, zero trust access control, spending limits, continuous monitoring, anomaly detection, revocation, audit trails, policy engines, and human approval for high-risk actions.

For higher-risk AI and Physical AI systems, controls may also include formal certification, independent review, testing, validation and verification, behavioural monitoring, risk scoring, incident reporting, safety-case evidence, safe-state mechanisms, and mechanisms to suspend or revoke machine credentials when safety, security, or compliance conditions are no longer met.

Current and emerging examples

Current examples are usually bounded systems, not fully open machine economies.

Electric vehicle charging uses machine-readable authentication and contract-based charging flows in systems such as ISO 15118 Plug & Charge.[19]

Energy flexibility and transactive energy use market-like mechanisms to coordinate distributed energy resources, demand response, batteries, and grid assets.[60][61]

Industrial data spaces use identity, credentials, policies, and connectors to support governed data exchange between organizations and technical systems. Gaia-X and Catena-X show how identity and verifiable credentials can support trusted data exchange.[29][30]

Programmable payments support use cases such as pay-per-use machinery, automatic charging payments, IoT payments, and conditional settlement.[48]

Societal implications

The machine economy may reduce transaction costs and enable more granular markets for data, energy, compute, mobility, logistics, and industrial services. It may also create risks of platform dependency, surveillance, market concentration, opaque automated decision-making, cyberattacks, labour displacement, regulatory gaps, unclear liability, and tax-base shifts.

Labour and ownership questions are central. Machines may create value, but the legal and economic ownership of that value is normally assigned to human owners, firms, operators, or platform providers. This can create principal-agent problems when autonomous agents act on behalf of users or companies.

The topic is also linked to debates about zero marginal cost and digital coordination. Jeremy Rifkin has argued that IoT and digital networks may reduce marginal costs in parts of the economy.[85] However, machine economy systems do not eliminate physical constraints. Machines still require energy, materials, maintenance, software updates, supervision, cybersecurity, insurance, taxation, and legal accountability.

Science fiction and cultural context

Science fiction has shaped public imagination about autonomous machines, robot labour, machine personhood, AI-managed societies, and post-scarcity economies. Karel Čapek's R.U.R., published in 1920 and first performed in 1921, introduced the word "robot" in the context of artificial workers.[86]

Isaac Asimov's robot stories are relevant to machine control, safety, and delegated action, especially through the cultural role of the Three Laws of Robotics.[87] Iain M. Banks's Culture novels are relevant to AI-governed and post-scarcity societies; academic work has used the Culture series to explore the political role of artificial intelligences.[88]

These works are cultural context. They are not evidence of technical feasibility.

Criticism and open questions

The term "machine economy" is broad and sometimes used in promotional contexts. Critics may argue that it combines several different technologies under one label without clear boundaries. A neutral encyclopedic treatment should therefore avoid presenting the machine economy as an inevitable future or a single technology trend.

Open questions include how machine authority should be represented, who may issue machine credentials, how revocation should work across ecosystems, how machine wallets should be governed, who is liable for autonomous transactions, how machine-generated value should be taxed, how regulators should supervise machine-to-machine markets, and how physical AI systems can be stopped or constrained safely.

Blockchain-related claims require special care. Blockchain can support some machine economy functions, but it does not by itself solve identity, legal authority, compliance, taxation, safety, or liability.

See also

References

Further reading

Related Articles

Wikiwand AI