Ethical escalation in artificial intelligence (AI)-enabled business leadership has become a central issue for business ethics because digitally globalized firms increasingly make consequential decisions through data infrastructures, platform rules, and vendor systems that operate across organizational and jurisdictional boundaries. Previous literature offers important principles for stakeholder responsibility, responsible innovation, and algorithmic accountability, yet it gives limited guidance on the exact point at which a legally permissible digital practice should be moved from ordinary managerial approval to documented ethical review. The purpose of this theoretical paper is to develop a new authorial concept that gives leaders a disciplined way to recognize, assess, and govern this escalation problem without reducing ethical judgment to compliance or technical scoring. The proposed Ethical Friction Threshold Model (EFTM) defines ethical friction as the ethically relevant tension created when digital practices cross boundaries of law, culture, organizational control, or stakeholder voice, and it combines boundary triggers, stakeholder exposure mapping, seven assessment dimensions, a boundary multiplier, an ethical threshold index, threshold classification, governance action, and learning records. The scientific novelty of the model lies in its explicit transformation of diffuse cross-boundary ethical tension into a transparent leadership threshold that links stakeholder voice, contestability, and value-chain responsibility to specific escalation decisions. Unlike principle catalogs, risk management standards, or audit checklists, the model identifies when ethical concern becomes strong enough to require explanation, redesign, pause, executive escalation, or board review while still preserving contextual judgment and written justification. The model reveals that ethical risk in digital globalization often grows not from a single severe feature but from the accumulation of moderate concerns across stakeholder exposure, power imbalance, jurisdictional distance, data opacity, decision irreversibility, weak contestability, and diffused responsibility. It also shows that learning from appeals, complaints, audits, and post-deployment monitoring should revise later trigger definitions, weighting choices, and safeguards so that accountability develops through repeated leadership decisions rather than a one-time approval event. The article opens a path for future case studies, expert panels, scenario experiments, and sector-specific governance tools in business ethics, digital leadership, marketing, procurement, finance, and platform management.