The development of artificial intelligence raises a question for Armenia not so much about its ability to create world-class technologies of its own as about the state's readiness to establish the conditions for their effective and safe use. This involves data quality, access to computing resources, training specialists, and the ability to control technologies on which government processes increasingly depend. COPA co-founder Henrik Sergoyan discussed the conditions needed to build a domestic AI ecosystem and the role Armenian developers can play in it at the 11th Armenian Internet Governance Forum, ArmIGF2026.

According to him, the question of whether Armenia needs to develop its own artificial intelligence should be considered in light of the tasks the state intends to address. When it comes to introducing AI into public administration and processing the most sensitive data, publicly available tools such as ChatGPT, Gemini, or Claude cannot be relied upon.

However, Sergoyan stressed that it is important to distinguish between creating an entirely new artificial intelligence model and adapting existing open models to specific tasks. “Very often, people confuse creating an entirely new AI with adapting open tools and AI-based models to their own needs,” he noted.

Sergoyan explained that COPA develops solutions that can run directly on a user's computer without an internet connection. According to him, this kind of AI can be compared to an ordinary computer game that launches and runs locally without requiring constant network access.

Data, Funding, and Computing Power

Several conditions are necessary to create sovereign AI. First and foremost, artificial intelligence requires data. That data must be accessible and of high quality. In this respect, Sergoyan believes Armenia already has the necessary institutional foundation: he pointed to the Information Systems Agency (ISA), which is responsible for improving data quality and ensuring data continuity within the public administration system.

The second factor is funding. According to Sergoyan, AI development requires substantial investment, including in specialists. His company employs eight scientists specializing in artificial intelligence. Their task is to create highly accurate and secure models that could be used in the public sector.

He also described the opening of the Firebird computing center as an important step. The state invested $25 million in the project, making provision for the use of its computing resources, including for public administration needs.

Sergoyan stressed that independently renting the necessary number of graphics processing units (GPUs) would have been a significantly more difficult task for private companies and the state itself. Now that the issue of computing capacity has been resolved, it is necessary to move on to the practical development and deployment of solutions.

Should the State Give Preference to Armenian Developers?

Sergoyan believes that the state should focus primarily on the transparency, cost, and quality of solutions. At the same time, he argues that foreign companies should not be automatically excluded: if developers registered in Estonia, Russia, or the United States offer a comparable level of quality, transparency, and soundness in their solutions, their proposals should also be considered.

At the same time, he pointed to Armenian companies' advantage in developing artificial intelligence for the Armenian language. According to him, the best results in this field are achieved within the country itself.

Sergoyan said that COPA places particular emphasis on openness in its development work and presents its work at international conferences. In particular, the company is researching ways to adapt AI for languages with limited resources—known as low-resource languages—which include Armenian. Its developers are looking for ways to adapt models even when only a small amount of language data is available.

At the same time, Sergoyan considers knowledge of Armenian desirable for specialists working in this field, but believes it should not become a mandatory requirement.

Sovereignty Means Control, Not Isolation

Speaking about technological sovereignty, Sergoyan supported the view that it does not mean having to produce absolutely every technology domestically. Sovereignty means being able to address challenges independently using existing developments, control the solutions being used, and retain domestic expertise.

According to him, Armenia should not set itself the goal of creating more advanced versions of GPT or Gemini: these products have already been developed by major international companies. The country's task is to make the most of existing open and transparent technologies by adapting them to its own needs.

In practical terms, this means that the state can use ready-made models, move them into its own infrastructure, connect them to government data, and create specialized solutions that work effectively with national datasets. At the same time, the data itself and the processes used to handle it must remain under the country's control.

Sergoyan described the governance of such systems as an equally important issue. In his view, humans must retain their role as the ultimate decision-makers, and artificial intelligence should not be granted independent authority, particularly when handling sensitive data in the public and private sectors.