AI tutors are boosting homework—but quietly eroding exam performance and reshaping security risk

Two new research threads are converging on the same uncomfortable question: what happens when generative AI becomes a default study and decision tool. A study cited by SCMP, based on 26,000 Chinese students, found that AI-assisted homework can raise homework scores while cutting exam results by about 20%. The “brain drain” effect is not immediate; the study indicates it takes roughly two years to fully emerge, implying a delayed erosion of test-time skills. In parallel, an IMF Selected Issues paper examines how AI is changing Israel’s labor market and productivity dynamics, focusing on workforce adjustment rather than classroom outcomes. Geopolitically, the classroom and the battlefield are starting to rhyme. If AI reduces the incentive or ability to practice under exam conditions, it can weaken human capital pipelines that governments and firms rely on for high-skill growth, which in turn affects long-run competitiveness and national security capacity. Meanwhile, the Just Security analysis frames military AI as a civilian protection challenge, highlighting how automation and AI-enabled targeting or decision support can alter risk to non-combatants even when intent is not explicitly hostile. The power dynamic is shifting toward actors that control data, models, and deployment pathways—governments, defense contractors, and large AI labs—while states with weaker governance or slower adoption face both productivity disruption and higher operational risk. Israel’s labor-market focus also matters because it signals how a small, innovation-driven economy may absorb AI-driven job reallocation pressures that can influence policy choices and social stability. Market implications cut across education, labor, and defense technology. In education-linked consumer and edtech segments, the reported 20% exam-score decline suggests demand may shift from “AI as a crutch” toward AI systems that enforce learning verification, tutoring transparency, and assessment integrity, potentially pressuring margins for low-integrity tools. For Israel, the IMF framing implies that AI-driven productivity gains may come with labor reallocation costs, which can affect domestic consumption, wage dynamics, and the risk premium on local equities tied to labor-intensive services. On the security side, the civilian-protection angle can translate into procurement and compliance costs for militaries and contractors—raising demand for explainable AI, human-in-the-loop architectures, and civilian harm mitigation tooling. While the articles do not name specific tickers, the likely beneficiaries are firms in AI governance, model evaluation, and defense AI safety, whereas higher-risk exposure sits with vendors whose systems lack auditability. The next watch points are whether policymakers treat these findings as separate education and defense issues or as one governance problem. For education, key indicators include longitudinal exam performance trends across cohorts using AI homework tools, plus regulatory moves on assessment design and AI disclosure in schools. For labor markets, investors should monitor Israel’s employment composition, participation rates, and productivity metrics as AI adoption accelerates, alongside any IMF-recommended policy adjustments. For military AI, the trigger is whether civilian-protection frameworks become binding procurement standards—e.g., requirements for testing, reporting, and operational constraints on automated decision support. Escalation risk would rise if AI deployment outpaces governance, while de-escalation would follow if states converge on verification and accountability norms for both education and defense systems.

Geopolitical Implications

• 01

  AI governance is becoming a cross-domain strategic issue: education integrity, labor-market stability, and civilian protection in military operations are converging under the same accountability theme.

• 02

  States and firms that control model access, evaluation, and deployment oversight gain leverage, while laggards face both human-capital erosion and higher operational risk.

• 03

  Procurement and compliance requirements for military AI could reshape defense contractor competition, shifting budgets toward safety, testing, and explainability.

Key Signals

• —Longitudinal exam-performance data from cohorts using AI homework tools, including whether regulators require disclosure or verification.
• —Israel labor-market indicators: employment composition, participation rates, and productivity trends as AI adoption scales.
• —Defense procurement language: explicit civilian-harm mitigation requirements, audit trails, and human-in-the-loop thresholds for AI decision support.
• —Any emerging international norms or reporting frameworks for military AI civilian protection.