PISA Finds One-Off AI Use Lowers Student Science Scores While Guided Use Helps

Overview

• The OECD's PISA 2025 data show students who use AI for one-off tasks score about 20 points lower in science, a gap comparable to nearly one year of schooling, while frequent use paired with school-taught digital literacy is linked to slightly above‑average results.
• In Brazil 48% of students report weekly chatbot use, yet schools score far below the OECD average on digital readiness, exposing how unequal access to devices, internet and teacher training can turn AI into a learning disadvantage.
• Education experts quoted in coverage warn that uncritical AI use can replace effort and harm learning and recommend teaching students to verify AI outputs and assess processes rather than imposing broad bans on access.
• Corporate studies show rapid growth in AI adoption but limited scale: a national survey found adoption rose to roughly 45% in 2026 while an industry benchmark reports only 15–19% of logistics firms use AI at scale, with talent shortages, poor data quality and integration complexity as the main barriers.
• Security reports describe AI‑augmented cyberattacks that used automated agents and the Anthropic model Claude to breach South Korean banks and leak customer data, a development that has prompted police inquiries and intensified calls for clearer governance of AI tools.