The data used: personal, employee, health, customer conversations, proprietary or third-party datasets.
Armenia. AI Regulation & Compliance
AI Regulation & Compliance in Armenia
Armenia has no single AI Act governing every system, but AI products do not operate outside the law. MOSTAR maps data, IP, contract, cybersecurity, liability and sector requirements and builds them into products and processes.
Discuss your AI productWhat determines the legal requirements
What AI does: generate, classify, recommend, predict, rank, support decisions or automate processes.
Who uses or is affected: consumers, employees, companies, patients, professionals or public-sector users.
The sector: SaaS, HR, healthcare, finance, cybersecurity, advertising or education.
Where users are located: foreign-market regulation may apply independently to an Armenian company.
AI legal and compliance services
AI regulatory mapping
Identify Armenian legal regimes and sector-specific requirements.
Data protection and AI
Review data flows, training data, prompts, logs, outputs, vendor access and transfers.
AI contracts
Draft development, SaaS, API/vendor and enterprise agreements, licences and data clauses.
Terms of Service
Structure permitted use, limitations, user obligations, outputs and remedies.
IP and training data
Review software, datasets, materials, outputs, third-party content and open source.
Vendor and API review
Assess dependencies involving models, APIs, cloud and external data.
Liability allocation
Allocate risk among developers, providers, customers and end users.
AI governance
Build approvals, human review, documentation, vendor controls and responsibility.
AI due diligence
Review ownership, data rights, vendors, licences and gaps before a transaction.
Regulatory monitoring
Track developments relevant to a specific product where included in scope.
When an AI legal review is useful
Launching an AI product
Identify the relevant legal regimes before public launch.
Using a third-party AI provider
OpenAI, Anthropic, Google, Microsoft or another provider becomes part of the product's data and contract chain.
Training or fine-tuning a model
Review data sources, rights, personal/confidential information and contractual restrictions.
Purchasing an AI development
Define ownership, access to code/models/data, licensing, maintenance, lock-in and liability.
AI affects decisions about people
HR screening, scoring and recommendations warrant closer review of data, transparency and liability.
Offering the product outside Armenia
Armenian law is one layer. Foreign AI, privacy and sector rules may apply independently.
How an AI compliance review works
Understand the use case
What the system does, who uses it, what outputs it creates and where it is offered.
Map data and technology
Identify data sources, processing locations, models, APIs, vendors and access.
Map applicable law
Identify Armenian regimes and where separate foreign-law review is required.
Identify gaps
Review data clauses, IP, user/vendor terms, liability, security and governance.
Implement
Prepare contracts, policies, product wording, procedures and risk allocation.
Key legal questions
Current regulatory landscape
No single horizontal Armenian law equivalent to the EU AI Act has been identified as a comprehensive framework for every AI system.
That does not mean AI is unregulated. The relevance of data, IP, contracts, cybersecurity, employment and sector rules depends on the use case.
AI and personal data
Personal data may appear in prompts, conversations, documents, logs, support tickets, datasets, retrieval systems, outputs and analytics.
Where a foreign provider accesses Armenian data, cross-border transfer requirements should be assessed.
Data protection in ArmeniaIP and training data
Review software, code, weights, datasets, third-party content, prompts, outputs, integrations and improvements separately.
Ownership and permitted use depend on human contribution, contracts, licences and the origin of materials.
Intellectual property in ArmeniaAI contracts and liability
An agreement may cover permitted use, performance, human review, data responsibility, IP, vendors, confidentiality, security, changes, liability and termination.
Contract clauses cannot override mandatory law, and disclaimers do not eliminate every legal responsibility.
Third-party models and APIs
Review data sent, training use, processing location, retention, confidentiality, enterprise terms, IP, restrictions, security, sub-processors and customer responsibility.
Using an external model does not remove the business's own legal risk.
Cybersecurity and AI
Armenia's Cybersecurity Law entered into force in January 2026 and establishes requirements for defined systems, critical infrastructure and providers.
Not every AI startup is automatically in scope. The analysis depends on the entity, service and infrastructure.
Foreign markets
Company incorporation and product-market regulation are separate questions. The EU, UK, US and other markets may impose additional rules.
MOSTAR can map the Armenian layer and flag where separate foreign-law analysis is required.
AI governance
Internal controls become useful where staff use public AI, prompts contain confidential data, multiple vendors are used, or AI affects decisions.
Possible controls include acceptable-use rules, vendor approval, data rules, human review, escalation, change documentation, incidents and responsibility mapping.
Frequently asked questions
Does Armenia have an AI Act?+
No single horizontal law equivalent to the EU AI Act has been identified as a comprehensive framework for every AI system. Other laws may apply to a specific product.
Does that mean AI is unregulated?+
No. Requirements depend on the use case and may involve data protection, IP, contracts, cybersecurity, liability and sector rules.
Does an AI startup need a special licence?+
A universal licence merely because a product uses AI should not be assumed. Healthcare, finance, security and other regulated activities may have separate requirements.
Can personal data be used for training?+
The source, purpose, legal basis, transparency or consent, access, retention, vendors and international transfers should be assessed.
Who owns AI-generated content?+
There is no universal rule. The output, human contribution, source materials, platform terms, contracts and applicable IP law matter.
What should an AI development agreement cover?+
Scope, deliverables, ownership or licensing, data use, third-party tools, confidentiality, security, vendor dependencies, acceptance, liability and post-termination rights.
Do we need new Terms of Service?+
If AI materially affects the service, generic SaaS terms may not be enough. Limitations, user obligations, data/IP, third-party models and risk allocation should be reviewed.
Does the EU AI Act apply to an Armenian company?+
Not automatically because it is incorporated in Armenia. EU rules may become relevant depending on where and how the system is offered, deployed or used.
Can Armenia's Cybersecurity Law apply to an AI company?+
Yes, where the company, system, service or infrastructure falls within its scope. Not every AI startup is automatically regulated.
Related services
Armenia. AI Regulation & Compliance
Building or deploying an AI product?
Describe the system, data, users and third-party models. We can identify the main legal questions and define the review scope.
We respond within 24 hours.

