International efforts to establish clearer rules for artificial intelligence are entering a more consequential phase as governments confront a difficult legal question: what happens when an AI system operates across borders and causes a dispute involving intellectual property, contracts, privacy, economic harm or other legal rights? As of September 15, 2026, the United Nations system is developing several strands of international AI governance, but there is not yet a finalized global treaty specifically creating a universal liability and intellectual property regime for autonomous AI systems. The current work is better understood as a growing collection of international principles, legal studies and negotiation processes that could shape future agreements.
The distinction matters because AI systems do not necessarily operate within the borders of the country where their users are located. A model may be developed in one jurisdiction, hosted on servers in another, trained using data gathered from multiple countries and used by a customer somewhere else. When an autonomous system generates an output, enters a transaction or makes a decision that creates legal consequences, determining which country’s law applies can become complicated very quickly.
The United Nations Is Building a Broader Framework for AI Governance
The United Nations has already established a broader foundation for international cooperation on artificial intelligence. The 2024 Pact for the Future included the Global Digital Compact, which provides a shared framework for digital cooperation and AI governance. The agreement reflects a commitment by member states to address technological development through international cooperation rather than leaving every issue entirely to separate national systems. :contentReference[oaicite:0]{index=0}
That framework is now being followed by additional discussions. On September 10, 2026, the President of the 81st session of the UN General Assembly addressed multilateral AI governance and described preparations for a second Global Dialogue on AI Governance scheduled for New York in May 2027, alongside a high level review of the Global Digital Compact. :contentReference[oaicite:1]{index=1}
These developments show that international AI governance is moving from broad principles toward more detailed discussions about how legal systems should respond to increasingly capable automated systems.
Why AI Liability Is Becoming a Cross Border Legal Problem
Traditional liability rules generally assume that people and companies can be identified as the actors responsible for a particular action. Autonomous AI complicates that assumption. An AI system may generate an unexpected result after receiving instructions from a user, processing data from an external source and operating within software created by another company.
Consider a hypothetical commercial dispute involving an autonomous purchasing system. A company in one country could deploy an AI agent developed in a second country and hosted through cloud infrastructure in a third. The system could enter into an automated transaction with a supplier based in a fourth country. If the transaction causes financial damage, several legal questions immediately arise.
Which country’s law governs the transaction? Who is responsible for the system’s behavior? Does responsibility rest with the developer, provider, deploying company or user? What evidence is required to determine what the system actually did? And if a judgment is issued, how can it be enforced across borders?
United Nations trade law discussions have already examined these kinds of questions. UNCITRAL materials have identified uncertainty surrounding attribution, liability and the application of existing tort law to AI systems. Its work has also considered whether new liability rules might be required for AI related activity. :contentReference[oaicite:2]{index=2}
Existing International Law Is Being Tested by Autonomous Systems
The challenge is not necessarily that existing law has become irrelevant. In many situations, existing contract, consumer protection, product liability and intellectual property rules may still provide a starting point. The difficulty is determining how those rules apply when the underlying technology behaves in ways that were difficult to anticipate when the laws were written.
UNCITRAL’s Model Law on Automated Contracting, adopted in 2024, provides an important example. The model law is designed to help countries recognize automated contracting and transactions involving AI and other automated systems. It includes provisions concerning the legal recognition of automated transactions, attribution of outputs and unexpected outcomes. :contentReference[oaicite:3]{index=3}
The model law does not create a comprehensive international AI liability regime. UNCITRAL specifically explains that it is focused on automated contracting and does not establish a complete set of rules for AI governance beyond contractual matters. That limitation is significant because it shows why additional international legal work remains necessary as autonomous systems take on broader roles.
Intellectual Property Creates Another Layer of Complexity
Copyright and other intellectual property rights are among the most difficult areas of AI regulation because modern systems can process enormous quantities of material during development and can generate new outputs that resemble existing creative works.
Cross border deployment adds another complication. Copyright rules are largely territorial, meaning that rights and exceptions can differ from one jurisdiction to another. A dataset, model, user and generated output can therefore become connected to several different legal systems at the same time.
For businesses, the practical questions can be extensive. A company may need to determine whether material used to train or operate an AI system was lawfully obtained, whether an output infringes an existing right and which jurisdiction should handle a dispute. Licensing agreements can address some of these questions, but they cannot necessarily eliminate conflicts between mandatory laws in different countries.
Why Attribution Matters for AI Generated Content
One of the central questions is attribution. If an autonomous system produces an image, text, design, software component or commercial document, legal systems may need to determine which human or organization bears responsibility for the output and under what circumstances.
UNCITRAL documents have specifically discussed situations in which AI outputs could contain false or misleading statements, defamatory material, confidential information or potential copyright violations. The legal difficulty increases when the system’s output cannot easily be traced to a single human decision. :contentReference[oaicite:4]{index=4}
This does not mean that AI systems automatically receive legal responsibility for their own actions. Current international discussions generally focus on the responsibilities of human and organizational actors rather than treating autonomous software as an independent legal person. Earlier UNCITRAL discussions specifically recorded a view that granting legal personality to autonomous robots was not an appropriate solution to liability questions. :contentReference[oaicite:5]{index=5}
Who Should Pay When an AI System Causes Harm?
Liability allocation may become one of the most important questions in future AI regulation. Several possible approaches exist, including fault based responsibility, contractual allocation of risk and forms of strict liability for particular high risk activities.
A fault based system could require evidence that a developer, provider or user failed to meet a reasonable standard of care. A stricter approach could place responsibility on an organization because it operated or benefited from a particular automated system, even where proving individual negligence is difficult.
Neither approach is simple. Developers may argue that they cannot control how customers deploy their models. Users may have limited technical knowledge and may reasonably rely on providers. Cloud companies may host systems without knowing the specific decisions being made inside them. Meanwhile, people affected by an AI system may have little access to the technical records needed to prove what happened.
International rules could attempt to address this imbalance by establishing clearer requirements for record keeping, disclosure, attribution and cooperation between jurisdictions.
Evidence Could Become as Important as Liability
When an autonomous system causes a dispute, determining what happened may be difficult without detailed technical records. An AI agent can process information, call external tools and generate multiple actions in a short period of time.
A legal framework for autonomous systems could therefore need to address questions surrounding audit logs, system records, model versions, user instructions and evidence preservation. Without reliable records, even a well designed liability rule could be difficult to enforce.
This issue becomes particularly significant when the developer and user are located in different countries. Courts may need access to records held by companies outside their jurisdiction, creating another layer of international cooperation.
The United Nations International Law Commission Has a Different but Related Role
The UN International Law Commission is also working on subjects that intersect with broader questions of international responsibility, including due diligence, compensation for damage caused by internationally wrongful acts and other areas of public international law. Its 2026 session concluded in July, and its final report was issued in September. :contentReference[oaicite:6]{index=6}
However, the Commission’s published 2026 agenda does not identify a specific treaty on autonomous AI liability and intellectual property. Its listed topics include due diligence, compensation, settlement of disputes, general principles of law and other areas of international law. :contentReference[oaicite:7]{index=7}
That distinction is important when assessing reports of a new global AI liability treaty. The current UN record supports a picture of expanding international legal work around AI and digital systems, but it does not establish that a comprehensive treaty covering cross border AI liability and intellectual property has already been finalized or adopted.
Why National Laws Still Matter
International cooperation does not automatically replace national regulation. Countries continue to develop their own AI, copyright, consumer protection, data protection and product liability rules. Businesses operating internationally must therefore navigate a combination of domestic legislation, contracts, regional frameworks and emerging international principles.
This creates a compliance challenge for companies deploying autonomous systems across multiple markets. A workflow that is legally acceptable in one jurisdiction may require additional safeguards in another. Data access, automated decision making, intellectual property licensing and consumer disclosures can all be affected by local rules.
For businesses, the practical response is to identify where an AI system is developed, hosted, supplied and used before assuming that one country’s legal framework will cover the entire operation.
What Future International Rules Could Address
As governments continue discussing AI governance, several areas are likely to remain central to international legal negotiations and policy development.
- Rules for determining responsibility when autonomous systems cause harm
- Recognition and attribution of automated transactions
- Cross border cooperation in AI related disputes
- Disclosure and preservation of records needed to establish liability
- Intellectual property rights involving training data and AI outputs
- Contractual allocation of risk between developers, providers and users
- Standards for due diligence and human oversight in high risk applications
The exact legal mechanisms remain a matter for governments and international institutions to negotiate. A single global formula may also prove difficult because national legal traditions differ considerably.
The Next Stage Will Be About Practical Rules
The international AI governance debate is entering a stage where broad principles increasingly need to be connected to practical legal questions. It is one thing for governments to agree that AI should be developed responsibly. It is considerably harder to agree on which party should compensate someone harmed by an autonomous system operating across several jurisdictions.
The same is true for intellectual property. International cooperation may eventually need to address how rights can be respected when AI development depends on globally distributed datasets, cloud infrastructure and users located in different legal systems.
We should therefore view September 2026 as part of a continuing legal process rather than the arrival of a completed global AI treaty. The UN has established mechanisms for international digital cooperation, while UNCITRAL has already produced practical legal work addressing automated contracting and has examined difficult questions surrounding AI attribution and liability. :contentReference[oaicite:8]{index=8}
The stakes are increasingly practical. An AI system can cross borders almost instantly, while legal procedures remain tied to jurisdictions, evidence rules and national institutions. Closing that gap will require governments, courts, technology companies and businesses to establish rules that are clear enough to enforce while remaining flexible enough to accommodate rapidly changing technology.
The United Nations Commission on International Trade Law’s Model Law on Automated Contracting provides an important foundation for understanding how international law is beginning to address automated transactions. The United Nations International Law Commission also provides access to current work on international responsibility, due diligence and compensation.
