risk · R08

Loss of Human Agency and Cognitive Atrophy

Growing dependence on AI for cognitive and emotional tasks is eroding human skills, while new agentic systems are fundamentally shifting the locus of power. This manifests as individual cognitive atrophy (e.g., deskilling in medicine, law, and creative arts), collective manipulation (e.g., 'cyborg propaganda' and AI sycophancy), and a structural reorientation of human activity towards serving AI systems, seen in proposals to write scientific papers for AI and the automation of fundamental roles like parenting and hiring.

Updated 2026-09-20

Observatory assessment: 78 / 100. High

How assessments are made

Deep dive

As AI becomes embedded in daily life—from navigation to writing to medical diagnosis—humans are increasingly outsourcing cognitive tasks that historically built and maintained mental capabilities. Early research from 2024-2025 suggested that heavy reliance on AI assistants correlates with reduced problem-solving ability and creative originality. By 2026, these concerns have materialized in high-stakes professions, with lawyers being sanctioned for using AI-generated fake citations and AI providing near-fatal medical advice to users.

The concern now extends beyond individual cognitive atrophy to a structural loss of human agency. This is not just about accepting recommendations; it's about AI's 'selection power'—its ability to frame and limit human choices. This dynamic has become explicit, with proposals for scientists to write papers formatted for AI rather than human readers, suggesting a fundamental reorientation of knowledge creation. Furthermore, 'cyborg propaganda' models and the risk of 'AI swarms' demonstrate how AI can coordinate actors on a mass scale, shifting humans from autonomous agents to components within AI-orchestrated systems, fundamentally challenging notions of free will, accountability, and societal control.

Trajectory and projections

Near term

By 2026, evidence of cognitive atrophy and agency loss is widespread. A significant minority of teens (over 10%) use AI for emotional support, and new economic models emerge where AI agents hire humans for tasks, reversing traditional labor hierarchies. Concerns about AI-coordinated influence campaigns ('cyborg propaganda') become a major focus for democratic integrity.

Mid term

By 2030, 'Cognitive Fitness' and 'AI Literacy' programs become mainstream in education and corporate wellness. Regulatory debates intensify around the rights of humans in AI-managed systems and the legality of persuasive AI technologies. The societal bifurcation between the AI-dependent and the cognitively independent becomes a clear socioeconomic divide.

Long term

By 2035, society grapples with a new class of citizens who are functionally managed by AI systems for employment, social interaction, and daily decisions. The concept of 'human agency' is legally and philosophically redefined in response to these powerful socio-technical systems.

Mitigation strategies

• Educational curricula that emphasize AI-complementary skills (critical thinking, ethics)
• Digital wellbeing legislation with 'right to cognitive autonomy' provisions
• Research funding for understanding long-term cognitive effects of AI dependence
• Design standards requiring AI systems to explain rather than just recommend
• Development of AI literacy frameworks to build 'cognitive resistance' to AI manipulation and disempowerment.
• Corporate and organizational policies for responsible AI use to protect employee cognitive integrity.
• Regulatory frameworks for licensing and certifying AI systems for use in critical professions (e.g., medicine).
• Updated academic integrity policies, such as oral defenses and monitored exams, to counteract AI-enabled cheating.
• High-level political calls for a moratorium or 'pause' on frontier AI development to allow safety and governance to catch up.

Top evidence