NewsBite analysis · AI, Business
YNetGlobal
Afraid AI will take your job? New study says not so fast
A study from Berkeley's Center for Responsible, Decentralized Intelligence found that no AI model surpassed a 24% success rate in completing complex professional tasks. The research examined over 1,500 tasks across various fields, revealing a significant gap between AI's advertised capabilities and actual performance. In contrast to industry claims of rapid progress towards autonomous systems, the findings suggest that jobs requiring critical thinking and adaptability may be safe for now. While automation will accelerate in routine tasks, complex roles will remain largely untouched in the near future.
MonitorConfidence: HighMid term horizonReputational risk: High
Published 21 Jul 2026, 20:10 UTC3 min read
Visit original article ↗Lead image — YNetGlobal
Summary
TL;DR
- Berkeley study tests AI on 1,500 real-world tasks
- No model exceeded a success rate of 24%
- AI struggles with tasks requiring continuous reasoning
- Findings challenge claims of AI's readiness for complex roles
- Regulatory caution in Europe contrasts with US tech promotion
- Impacts likely on routine, repetitive jobs rather than complex professions
Interesting facts
- No AI model managed to pass the 25% success threshold across assigned tasks.
- All tested models recorded a zero percent success rate on the most complex tasks.