1. Introduction

The rapid digitisation of electricity markets has created a new category of contracting problem: contracts formed, negotiated, accepted, priced, or performed by machines rather than directly by human beings. In energy markets, software agents can monitor electricity prices, forecast demand, submit bids, purchase renewable electricity, trigger battery charging or discharging, execute demand-response transactions, and manage power purchase agreements (PPAs).

A machine agent may be understood as software or an automated system authorised by a person or organisation to perform contractual functions. It may range from a simple pre-programmed trading algorithm to an AI system capable of making dynamic decisions.

The central legal question is not necessarily whether the machine itself should become a legal person. The more immediate question is:

When a machine generates or accepts a contractual offer, whose act is it, and when should that act create legal obligations for the human or organisation behind the system?

Modern electronic-commerce law increasingly answers this through attribution and automated-contracting rules rather than by granting machines independent legal personality. The UNCITRAL Model Law on Automated Contracting (2024), for example, expressly addresses AI, smart contracts and machine-to-machine transactions and provides rules concerning automated outputs and their attribution to persons. UNCITRAL

2. Meaning of Machine Agents in Energy Contracts

A machine agent is an automated technological system that performs contractual or commercially significant actions on behalf of a human or legal entity.

In the energy sector, examples include:

• an electricity-trading algorithm submitting bids to a power exchange;
• an AI system selecting the cheapest electricity supplier;
• a battery-management system automatically purchasing electricity;
• an aggregator's software enrolling consumers into demand-response programmes;
• an automated system executing renewable-energy certificates;
• an AI system negotiating electricity prices;
• a smart-grid controller triggering electricity purchases;
• software automatically accepting balancing-market offers;
• autonomous EV-charging systems responding to electricity prices; and
• smart contracts automatically releasing payment after delivery of electricity.

The machine therefore functions operationally as an interface between the principal and the energy market.

Importantly, this does not automatically mean that the machine has legal personality.

3. Machine Agency Versus Legal Personality

A critical distinction must be maintained between acting as an agent and being a legal person.

Traditional agency law normally involves:

Principal → Agent → Third Party

The agent acts within authority and creates legal consequences for the principal.

With an AI or machine agent, the structure becomes:

Principal → AI/Machine System → Energy Market Counterparty

The machine may perform the operational function of an agent, but the legal rights and obligations normally remain attached to the human or corporate principal.

This distinction is particularly important because current legal frameworks generally do not require AI systems to possess independent contractual personality merely because they participate in contract formation.

The UNCITRAL Model Law on Automated Contracting deliberately approaches the problem through recognition of automated contracting rather than by creating a new legal personality for AI. UNCITRAL

Thus:

Legal recognition of machine-generated contracts does not necessarily require legal recognition of machines as legal persons.

4. Evolution of the Legal Concept

Traditional contract law developed around human communication.

A conventional contract involves:

1. offer;
2. acceptance;
3. intention to create legal relations;
4. consideration where required;
5. capacity;
6. certainty; and
7. compliance with formal requirements.

Automation complicates the concept of acceptance because there may be no human examining the individual transaction.

For example:

A battery-management algorithm observes that electricity is available at ₹3/kWh, determines that charging is economically optimal, and automatically purchases 5 MWh.

No employee may have personally reviewed the transaction.

The legal question becomes whether the automated purchase is nevertheless attributable to the company that deployed the algorithm.

Modern electronic-commerce legislation has increasingly recognised this possibility.

5. UNCITRAL Model Law on Automated Contracting 2024

The most significant international development is the UNCITRAL Model Law on Automated Contracting (MLAC), adopted on 11 July 2024.

It specifically addresses:

• automated systems;
• AI techniques;
• smart contracts;
• machine-to-machine transactions;
• automated contract formation;
• automated contractual performance;
• attribution of automated outputs;
• computer code;
• dynamic information; and
• unexpected outcomes.

UNCITRAL explains that the Model Law is designed to provide legal certainty for automated contracting without requiring a separate legal regime for every technology. UNCITRAL

It follows three important principles:

A. Technology neutrality

The law should not depend on whether the machine uses:

• AI,
• blockchain,
• conventional software,
• smart contracts,
• machine learning, or
• another technology.

B. Non-discrimination

A contract should not lose legal validity merely because automation was used.

C. Party autonomy

Parties should generally be able to decide whether and how automated systems will be used in their contractual relationship, subject to mandatory law. UNCITRAL

These principles are particularly suitable for electricity markets, where transactions occur at high speed and may involve thousands of automated decisions.

6. Earlier UNCITRAL Recognition of Automated Transactions

The concept is not entirely new.

The UNCITRAL Model Law on Electronic Commerce 1996 established the principles of non-discrimination, technological neutrality and functional equivalence for electronic transactions. UNCITRAL

The United Nations Convention on the Use of Electronic Communications in International Contracts 2005 went further.

Article 12 addresses contracts formed through automated message systems and makes clear that the absence of human review of an individual transaction does not, by itself, prevent contractual formation. UNCITRAL

This principle is extremely important for energy markets.

Suppose an electricity trader authorises software to submit bids every five minutes. The trader does not personally approve every bid. The absence of individual human review should not automatically invalidate each transaction.

The legal focus instead shifts to:

• authorisation;
• system configuration;
• contractual terms;
• attribution;
• authentication;
• error;
• fraud;
• scope of authority; and
• applicable market rules.

7. Indian Legal Position

India currently has no comprehensive statutory framework specifically creating independent legal personality for AI agents. Current legal treatment is better understood through existing contract and electronic-transactions law.

The principal legislation includes:

• Indian Contract Act, 1872;
• Information Technology Act, 2000;
• Indian Evidence Act framework as modified by electronic-evidence legislation;
• Bharatiya Sakshya Adhiniyam, 2023;
• Arbitration and Conciliation Act, 1996;
• sector-specific electricity legislation; and
• rules and regulations governing electricity markets.

The Information Technology Act is particularly important because electronic contracts cannot be rejected merely because electronic means were used.

Section 10A of the Information Technology Act recognises contracts formed through electronic means.

Therefore, the legal question for an AI-generated energy contract is generally not:

"Did a human physically click the acceptance button?"

It is more appropriately:

"Was the electronic action attributable to the party, authorised by it, and otherwise compliant with contract and sectoral law?"

8. Trimex International FZE Ltd. v. Vedanta Aluminium Ltd.

A major Indian authority is Trimex International FZE Ltd. v. Vedanta Aluminium Ltd., (2010) 3 SCC 1.

The dispute concerned a series of email communications relating to the supply of bauxite. The parties exchanged commercial terms electronically, and the Supreme Court considered whether a binding contract and arbitration agreement had arisen. Indian Kanoon

The Supreme Court held, in substance, that a contract could be concluded through electronic communications where the essential contractual requirements were satisfied.

The Court examined:

• offer;
• acceptance;
• intention;
• contractual terms;
• communication of acceptance; and
• the parties' conduct.

It found that the unconditional acceptance communicated by email resulted in a concluded contractual arrangement. Indian Kanoon

Relevance to machine agents

Trimex did not concern AI agents. Nevertheless, it establishes an important legal foundation:

The contractual validity of an agreement does not depend upon the use of paper or face-to-face communication.

Accordingly, where an energy company's authorised trading algorithm communicates an offer or acceptance electronically, the electronic nature of the communication should not by itself invalidate the transaction.

9. P.R. Transport Agency v. Union of India

In P.R. Transport Agency v. Union of India, AIR 2006 All 23, the Allahabad High Court considered issues concerning electronically communicated contractual arrangements and jurisdiction.

The case is significant in the development of Indian jurisprudence recognising the legal relevance of electronic communications in contractual relationships. Indian Kanoon

For machine-agent contracts, such jurisprudence supports a broader proposition:

electronic communication can produce legal consequences even where the contracting process does not follow traditional paper-based methods.

10. Quoine Pte Ltd v. B2C2 Ltd

One of the most directly relevant international cases is Quoine Pte Ltd v. B2C2 Ltd [2020] SGCA(I) 2, decided by the Singapore Court of Appeal.

This case involved cryptocurrency trading rather than electricity, but its legal reasoning is highly relevant to algorithmic energy markets.

B2C2 used trading software that operated automatically and concluded trades without direct human intervention. Due to a problem affecting the exchange's access to external market information, trades were executed at highly abnormal prices. Default

The court therefore confronted a fundamental question:

Can conventional contractual principles apply when transactions are formed entirely through algorithms?

The case demonstrates that automated contracting can be analysed through ordinary contract principles, while also raising difficult questions concerning:

• unilateral mistake;
• knowledge;
• intention;
• algorithmic design;
• attribution;
• contractual terms; and
• responsibility for abnormal automated outcomes.

Importantly, the trading software involved was deterministic: it operated according to programmed instructions rather than possessing independent legal personality. laws.sg

Energy-law significance

Consider a power-market algorithm that accidentally interprets a price signal incorrectly and submits electricity bids at an extreme price.

Quoine illustrates why an energy contract should not automatically be treated as invalid merely because a machine produced the transaction. Instead, courts may need to examine the applicable contractual rules, authority, system design, knowledge, error and market rules.

11. Register.com v. Verio

In Register.com, Inc. v. Verio, Inc., 356 F.3d 393 (2d Cir. 2004), automated software was used to make repeated queries to an online database.

The litigation concerned whether conduct performed through automated software could have contractual and other legal consequences.

The Second Circuit considered whether the repeated automated conduct manifested assent to terms governing access to the database. Justia Law

The case is relevant because it demonstrates that courts can analyse machine-generated conduct through ordinary doctrines of assent and contractual obligation, rather than treating the machine as an independent legal person.

For energy markets, a comparable question could arise where an automated trading system repeatedly interacts with a power exchange under published market rules.

12. Attribution: The Central Legal Principle

The most important issue is attribution.

Suppose an AI trading agent enters into a ₹50 crore electricity purchase agreement.

Who is legally bound?

Usually the relevant legal chain is:

AI output → authorised system → deploying organisation → contractual obligation

The machine itself generally does not become the debtor or creditor.

The UNCITRAL Model Law on Automated Contracting specifically develops rules dealing with attribution of automated-system outputs. UNCITRAL

This approach is particularly practical because it avoids creating thousands of artificial legal persons for:

• trading bots;
• battery controllers;
• smart meters;
• virtual power plants;
• energy-management systems.

13. Authority of the Machine Agent

Authority becomes a central contractual issue.

An energy company might authorise an AI agent to:

• purchase electricity up to ₹10/kWh;
• trade up to 10 MW;
• enter contracts for up to 24 hours;
• participate in balancing markets;
• purchase renewable power;
• sell stored electricity.

If the agent acts within those parameters, attribution is relatively straightforward.

The difficult situation occurs when the AI exceeds its authority.

For example:

The company authorises an algorithm to purchase electricity up to ₹10/kWh. Due to a software error, it purchases electricity at ₹100/kWh.

The legal questions include:

1. Was the algorithm actually authorised?
2. Were the limits communicated to counterparties?
3. Were the limits contractual or merely internal?
4. Was the counterparty entitled to rely on the machine's output?
5. Was there an obvious error?
6. Did the company's conduct create apparent authority?
7. Do electricity-market rules contain special error provisions?

14. Capacity of the Machine

Under conventional contract law, a machine does not ordinarily possess human legal capacity.

It cannot independently:

• own property;
• bear contractual obligations in its own name;
• sue or be sued merely because it is software;
• possess independent assets; or
• become a legal person simply through autonomous operation.

Consequently, machine agency should be distinguished from machine personality.

The better legal model is:

The machine performs the contractual function; the human or legal entity bears the legal rights and obligations.

15. Energy Contracts Particularly Suitable for Machine Agents

Machine agents have substantial potential in several energy-contract categories.

A. Power Purchase Agreements

AI systems can monitor:

• generation;
• weather;
• demand;
• prices;
• transmission constraints; and
• renewable output.

They can trigger contractual mechanisms such as volume adjustments or settlement calculations.

B. Electricity Trading

Algorithms can automatically:

• submit bids;
• accept offers;
• hedge positions;
• buy balancing energy;
• sell surplus power.

C. Demand Response

An automated system may reduce consumption when market prices or grid conditions reach specified thresholds.

D. Battery Contracts

Battery-management software can automatically decide:

charge → store → discharge → sell.

E. Peer-to-Peer Energy Trading

Smart contracts and AI agents may facilitate transactions among:

• households;
• rooftop solar producers;
• batteries;
• EVs; and
• local energy communities.

F. Renewable Energy Certificates

Automated systems may track production and automatically purchase or transfer certificates.

16. Smart Contracts and Machine Agents

Machine agents increasingly interact with smart contracts.

A smart contract may automatically execute:

"If electricity delivered ≥ 95% of contracted quantity, release payment."

An AI agent might supply the relevant data.

This creates a layered legal structure:

Underlying legal contract
↓
Computer code
↓
Machine agent
↓
Energy data
↓
Automated performance/payment

The legal contract and the software code should therefore not automatically be treated as identical.

UNCITRAL's 2024 Model Law expressly recognises the contractual significance of computer code and dynamic information in automated contracting. UNCITRAL

17. Error in Machine-Generated Energy Contracts

Errors are particularly important in electricity markets because transactions may occur in milliseconds.

Possible errors include:

• wrong price;
• wrong quantity;
• wrong delivery location;
• incorrect bidding algorithm;
• faulty meter data;
• cybersecurity intrusion;
• corrupted weather data;
• API failure;
• erroneous forecast;
• software bug; or
• malicious manipulation.

Traditional contract law may provide doctrines concerning:

• mistake;
• misrepresentation;
• fraud;
• duress;
• impossibility;
• frustration;
• breach; and
• restitution.

But market-specific regulations may provide additional cancellation or error-management mechanisms.

The UNCITRAL framework specifically recognises that automated systems create questions concerning unexpected outcomes and attribution. UNCITRAL

18. Cybersecurity and Machine-Agent Contracts

Cybersecurity creates another layer of legal responsibility.

Suppose a hacker gains control over an electricity-trading agent and causes it to purchase enormous quantities of electricity.

Potential questions include:

• Was the system adequately secured?
• Who controlled the credentials?
• Was multi-factor authentication required?
• Was the transaction cryptographically authenticated?
• Was the counterparty entitled to rely on the transaction?
• Did the attacker exceed the agent's authority?
• Who bears the market loss?

Therefore, an energy contract involving an autonomous machine should contain clear cybersecurity obligations.

19. Evidence and Auditability

Machine-agent contracts create enormous quantities of digital evidence.

Relevant evidence may include:

• source code;
• algorithm versions;
• system logs;
• API records;
• timestamps;
• authentication records;
• blockchain entries;
• market orders;
• meter data;
• model outputs;
• decision logs; and
• communications with counterparties.

This makes auditability a central principle of energy-law governance.

A party seeking to enforce a machine-generated contract may need to establish:

which system acted, under whose authority, using which rules, at what time, and based on which data.

20. Disclosure and Transparency

A sophisticated legal framework should require contractual disclosure concerning automated systems.

An energy contract could specify:

• that automated agents may form transactions;
• the limits of their authority;
• permitted transaction sizes;
• pricing boundaries;
• applicable algorithms;
• authentication procedures;
• error-correction procedures;
• suspension rights;
• audit rights;
• cybersecurity standards; and
• dispute-resolution mechanisms.

UNCITRAL's 2024 Model Law recognises the importance of disclosure concerning automated systems and states that automation cannot simply be used to evade other legal requirements. UNCITRAL

21. Consumer Protection

Machine contracting becomes more sensitive where consumers are involved.

Consider a household with an AI energy-management system.

The system may automatically:

• switch suppliers;
• change tariffs;
• participate in demand response;
• sell rooftop electricity;
• charge an EV;
• discharge a home battery.

Consumer law may impose mandatory protections that cannot simply be bypassed by saying:

"The AI agreed to the contract."

The legal responsibility remains with the relevant supplier, aggregator, platform or other regulated entity.

This is particularly important for vulnerable energy consumers, where automated decision-making could potentially affect access to essential electricity services.

22. Regulatory Responsibility

Energy is not an ordinary commercial commodity. Electricity is generally treated as a highly regulated essential service.

Consequently, machine-agent contracting must comply with sector-specific requirements concerning:

• licensing;
• electricity trading;
• market access;
• grid operation;
• tariff regulation;
• consumer protection;
• balancing;
• metering;
• renewable-energy obligations;
• cybersecurity;
• data protection; and
• system reliability.

An AI agent cannot create legal authority that its principal does not possess.

For example, if a company is not legally permitted to participate in a particular electricity market, its AI agent cannot acquire that permission merely by submitting a valid-looking electronic bid.

23. Legal Liability

A useful liability model is:

Developer

Potential responsibility for:

• defective software;
• negligent design;
• security vulnerabilities;
• misleading representations.

Deploying company

Potential responsibility for:

• authorisation;
• supervision;
• configuration;
• compliance;
• contractual performance.

Operator/User

Potential responsibility for:

• improper instructions;
• misuse;
• failure to monitor;
• ignoring warnings.

Energy-market operator

Potential responsibility where its own systems or market rules cause the failure.

Machine itself

Normally, the machine is the instrument through which conduct occurs, rather than an independent bearer of liability.

24. Proposed Legal Framework for Energy Machine Agents

A future energy-law framework could recognise six elements.

1. Legal recognition

Contracts should not be invalid solely because an automated agent participated in their formation.

2. Attribution

The output should be legally attributable to the person or organisation that authorised the system.

3. Authority limits

The principal should establish:

• financial limits;
• quantity limits;
• geographic limits;
• temporal limits; and
• permissible counterparties.

4. Auditability

Energy-market participants should maintain reliable transaction records.

5. Error correction

Rules should address:

• technical failures;
• erroneous data;
• cybersecurity attacks;
• abnormal bids; and
• unexpected AI decisions.

6. Human/legal accountability

Autonomous operation should not eliminate legal responsibility.

25. Case-Law-Based Legal Principles

26. Major Legal Challenges

Despite increasing recognition of automated contracting, several unresolved questions remain.

A. Autonomous decision-making

Traditional agency doctrine assumes an identifiable human or corporate principal. AI systems can make decisions that were not expressly anticipated by their developers.

B. Unforeseeable outcomes

An AI agent may select a contractual outcome that its creator did not specifically predict.

C. Algorithmic mistakes

Determining when an algorithmic error should invalidate a transaction remains difficult.

D. Cross-border transactions

Energy markets increasingly involve cross-border electricity trading and internationally supplied software.

Different jurisdictions may have different rules on:

• electronic signatures;
• automated contracts;
• AI liability;
• evidence;
• consumer protection; and
• energy regulation.

E. Explainability

A machine-learning system may be unable to provide a simple explanation of why it selected a particular transaction.

27. Future of Machine Agents in Energy Law

The development of AI-driven electricity markets suggests that machine agents will increasingly become part of the operational architecture of energy contracting.

Future systems may involve:

AI forecast agent → trading agent → grid agent → battery agent → settlement agent

Each system may interact with another automatically.

This could produce genuinely machine-to-machine energy markets.

UNCITRAL's 2024 Model Law is significant precisely because it recognises that automated contracting now extends beyond simple pre-programmed electronic transactions to AI-enabled and machine-to-machine contracting. UNCITRAL

However, legal recognition should not be confused with granting AI independent legal personality.

The more practical legal model is:

Recognise the contractual effect of machine actions while maintaining human or organisational accountability for those actions.

28. Conclusion

Legal recognition of machine agents in energy contracts represents a transition from human-centred contracting to technologically mediated contracting.

Existing electronic-commerce law already provides a substantial foundation. Indian law, particularly through the Information Technology Act and judicial recognition of electronic contracts in Trimex International v. Vedanta Aluminium, demonstrates that contractual validity does not necessarily depend upon traditional paper-based communication. Indian Kanoon

Internationally, the development has moved further. The UNCITRAL Model Law on Automated Contracting 2024 expressly addresses AI, smart contracts and machine-to-machine transactions and provides a framework based on technological neutrality, attribution and party autonomy. UNCITRAL

Cases such as Quoine v. B2C2 demonstrate the practical contractual difficulties created by algorithmic transactions, particularly concerning mistakes and unexpected outcomes. Default

For energy law, therefore, the emerging principle can be stated as follows:

A machine need not be a legal person to participate effectively in contract formation. The decisive legal issue is whether its actions can be validly attributed to an authorised human or legal entity and whether the resulting transaction complies with ordinary contract law and mandatory energy-market regulation.

This approach allows electricity markets to benefit from automation while preserving accountability, consumer protection, market integrity and regulatory control.