Researchers and technology analysts are identifying a scenario termed "gradual disempowerment," where the increasing use of artificial intelligence (AI) leads to the systematic displacement of human labor, decision-making, and cognitive skills. Unlike "rogue AI" scenarios often depicted in fiction, this process involves humans voluntarily ceding control to automated systems due to economic and political incentives.
The concept gained prominence following a paper published in early 2025. According to Jan Kulveit, co-founder of ACS Research Group and a co-author of the paper, the transition is driven by the perceived value AI brings to state, cultural, and economic functions rather than a deliberate power grab by machines. The shift began to accelerate following the 2022 launch of ChatGPT, as AI models demonstrated growing proficiency in cognitive tasks such as coding, math, and writing.
Jay Chooi, CEO of Robocurve, stated that gradual disempowerment ranks at the top of his concerns, comparing its potential impact to civilization-scale threats like engineered pandemics or cyberattacks. He noted that organizations face strong competitive pressure to phase out human workers in favor of AI agents and robots that do not require sleep, wages, or sick days. Margaret Mitchell, a research scientist at Hugging Face, added that when humans stop practicing skills outsourced to AI, they lose the ability to oversee or fix those systems if they fail.
The potential effects extend to governance and warfare. Kulveit argued that if states no longer rely on citizens for labor, taxes, or military service, they may move away from democratic models toward authoritarianism, similar to modern "oil states" that do not depend on a broad tax base. In military contexts, the pressure to use AI for faster decision-making could force all parties to remove humans from the loop to remain competitive, further reducing human agency.
For the individual, the change would be noticed in the gradual erosion of professional autonomy and the deterioration of specialized skills. A person might see their daily tasks automated, eventually leading to a loss of the expertise required to manage those tasks independently. This creates a cycle of dependency where humans become less capable of intervening when automated systems make errors. In the broader societal sense, this could lead to a loss of political leverage for citizens, as governments that rely on automated labor and AI-driven militaries may feel less accountability to their populations.
To address these risks, researchers advocate for "strategic friction," which involves intentionally designing AI to keep humans involved at critical steps. This would require shift from training AI as a human replacement to training it as a complementary tool that explains its reasoning and supports human education. While some tech leaders and ethicists are pushing for these proactive design changes, it is not yet known if or when such standards will be adopted across the industry. Currently, the primary focus remains on increasing AI capability and performance benchmarks.