An AI agent is a virtual assistant. It helps with tasks such as analyzing emails or making payments. Estonia wants to be the first country to regulate AI agents. But experts see no point in this.
An AI agent is not like ChatGPT. The agent receives a task and finds its own way to complete it. For example, a robot vacuum cleaner collects dust, but we don’t know exactly how it does it. People use agents to make life easier. For example, they make payments, analyze bank accounts, and phone bills.
In the world, agents are also used to buy plane tickets. The agent finds the cheapest ticket and buys it. This is dangerous because the agent uses a person’s credit card. One e-resident uses agents for tax offices and bank accounts without visiting their websites.
Estonia wants to give AI agents a unique code called the agent code. This helps distinguish humans from agents. Initially, agents were to be registered for both individuals and companies, but now only private individuals can do this. All codes go into one database where it can be checked who uses the agents.
The IT company Cybernetica believes this plan is dangerous. They wrote 11 pages of feedback and said the draft text is inadequate. Due to security risks, the draft should not be adopted. The security level must be as high as on an ID card.
Legal experts believe the law was changed too quickly. IT experts say technical issues were not considered. The Competition Authority does not support the draft because economic risks were not assessed. The Data Protection Inspectorate says there is no risk analysis.
ITL President Ats Albre says the country’s direction is right, but responsibility must be clear. He compares it to a robot vacuum cleaner: if a dog makes a mess and the robot spreads it, the person is responsible. Albre believes not everything needs to be regulated.
Minister Liina Vahtras says the plan is an opportunity, not an obligation. If the technical working group finds it impossible to implement, nothing will happen. The costs to the state are small.