Deep-sea polymetallic nodule (PMN) mining is emerging as a potential source of critical minerals for global energy and technology supply chains. Yet it operates under extreme environmental uncertainty and limited operational experience, exposing projects to substantial business interruption (BI) risk. This paper develops a probabilistic and economically interpretable framework for modeling aggregate BI loss distributions in PMN mining under data-scarce conditions. To our knowledge, no existing study provides a system-level actuarial loss model for PMN mining that integrates causal dependency, restoration dynamics, and economic valuation. The proposed approach combines a Bayesian Belief Network (BBN) with expert-elicited conditional probabilities. Results show that interruption probability ranges from 19.3% under favorable conditions to 77% under adverse regimes. Sensitivity analysis identifies the basic control system, structural integrity, buoyancy system, mining plan, and emergency response as dominant contributors to interruption probability. However, cost-based analysis reveals that the highest BI losses are driven by a smaller subset of technical components. For single interruptions, expected restoration times range from 17.5 ± 3.8 days to 52.7 ± 20.4 days. However, multiple interruptions may occur randomly within a year. Accounting for these conditions, insurance pricing for annual premiums supporting system recovery can vary from $1.1 M to $10.9 M.