ChatGPT and Anthropic amplified the "AI veganism" trend: why fatigue with neural networks is growing

AI became a part of everyday work and communication faster than trust in it could develop. Against the backdrop of ChatGPT's growing audience and the mass adoption of neural networks, an increasingly visible movement of users who want to limit AI in work, study, and even personal life is emerging.

Why the rejection emerged

According to research by NAFI and Ingosstrakh, 43% of Russians are ready to abandon artificial intelligence forever. Against this background, the term "AI-veganism" is increasingly heard — a deliberate limitation of neural networks in everyday life, work, and study. The logic here is roughly the same as digital detox: technologies are useful, but their excess begins to weigh.

The paradox is that this is happening not at the market's periphery, but at a moment when ChatGPT is approaching an audience of nearly a billion weekly users and is becoming a familiar household tool. The rejection is fueled not only by fatigue but also by fear for careers. Companies are actively implementing AI in development, support, and analytics, and around the market, conversations about layoffs and replacing specialists with agents are not dying down.

At the same time, the picture remains contradictory: some teams accelerate work through neural networks, others return people to roles they tried to automate, and in parallel, new professions emerge — from AI trainers to ethics specialists. For users, this looks not like a clear strategy but like an experiment with unpredictable consequences.

Benefit without magic

Despite the anxiety, demand for AI is only growing. In an Anthropic study involving over 80,000 Claude users from 159 countries, people most often said they expect professional development, personal transformation, and help managing their own lives from such systems. In practice, the picture is more grounded: AI best meets expectations where productivity growth is needed, quick structuring of thoughts, and support in learning.

At the same time, nearly one in five participants still admitted that the neural network did not give the result they expected. This is the main gap between AI's marketing image and real experience. Users want not a miracle machine, but an understandable assistant that speeds up routine, helps analyze data, and offers draft solutions, but doesn't pretend to be an infallible expert.

So far, the best use cases remain quite practical: write a first draft, gather conclusions from an array of information, highlight ideas, or help understand a new topic. As soon as AI starts portraying itself as a universal authority, trust quickly declines.

Where the anxiety comes from

Monitoring by the Nauka TV channel and the MOMRI institute shows that artificial intelligence has been the leading source of anxiety about scientific progress for three years in a row. By the end of 2025, the share of Russians frightened by AI and neural network development has grown from 15% to 27%. People are concerned not only with the speed of change itself, but also with the sense that the rules of the game are not yet determined: systems are becoming more accessible and influential faster than clear control over their quality and application emerges.

"The top AI-related fears are led by unreliability."

• Hallucinations and errors: research by Columbia Journalism Review showed that AI-powered search systems often incorrectly cite news sources.
• Risk of displacement from the labor market: fears of automation are amplified by loud announcements from companies and industry leaders.
• Phishing and deepfakes: according to Microsoft, users click on automated AI-phishing emails significantly more often than on emails written by people.
• False support: researchers at HSE in St. Petersburg found that language models inadequately respond to requests from people who need psychological help in more than 20% of cases.

A separate risk zone is emotional dependence on chatbots. When AI is used as a "psychologist," a conversational partner for reconciliation in relationships, or a source of life decisions, safety, not convenience, comes to the fore. A neural network can normalize dangerous symptoms, offer manipulative communication strategies, or simply confidently recommend an unsuitable solution. This is why skepticism about AI today is connected not only to work, but also to an attempt to prevent the algorithm from entering too sensitive areas of human life.

What this means

"AI-veganism" does not look like a mass rejection of technology, but rather an understandable reaction to inflated expectations and real risk of errors. Most likely, the market will not arrive at total adoption or prohibition, but rather at a more rigid selective use of AI: where it saves time and helps thinking, it will be implemented even faster, while in matters of money, health, safety, and emotions, demand for human verification will only grow.