Kirill Chernovol, Researcher of the International Best Practices Analysis Department at the Gaidar Institute, explained how concerns regarding artificial intelligence evolve as technology advances, and why issues of responsibility and control remain central.
According to the economist, a lack of trust in AI is common across various countries, particularly in areas where AI-driven decisions can impact people’s lives. For instance, the OECD data indicates that only four out of ten people trust AI-based decisions in public administration or believe that AI could help improve functioning of government agencies.
"In a broader sense, there are concerns that are always raised regarding the use of AI: namely, lack of control and anyone responsible for decisions AI makes. For instance, a company might automate tasks using AI, but it cannot punish or fire an AI for making mistakes. However, if you hire someone specifically to monitor the AI’s work, why automate the process, when that same person could just as easily perform the task?" explained Kirill Chernovol.
He noted that as technologies evolve, so do the primary concerns of users. While the initial adoption of AI in fields like finance, medicine, and transport centered on issues such as algorithmic transparency, potential discrimination, and protection of personal data, the spread of generative models has brought new concerns to the forefront, i.e., errors, fabricated information, and copyright infringement.
AI agents capable of independently interacting with other systems and accessing data are becoming increasingly widespread. Consequently, risks such as leakage of confidential information, unauthorized data deletion, and implementation of actions beyond the scope of the assigned task are becoming particularly critical.