Formal Board Control vs. Effective Operational Control

The company chop and the WFOE's registered legal representative, the two instruments that actually authorize contracts, banking transactions, and regulatory filings in China, sat with a single co-founder, outside any board approval process. Day-to-day control, including the ability to open or move bank accounts, execute vendor and customer contracts, and file regulatory paperwork, remained entirely with that individual regardless of what the board voted.

This was not a drafting oversight; the shareholder agreement's board-approval requirements were correctly worded and would have been enforceable in a jurisdiction where the signature, not the chop, is the binding act. The gap existed precisely because the legal documentation had been drafted against a Western governance template without a parallel chop-custody and legal-representative-succession mechanism layered on top of it — a pattern common enough that the lead investor's own standard diligence checklist did not include a specific line item for it.

Tracing the practical implication further: in a scenario where the board voted to remove or restrict that co-founder's authority, the co-founder would retain the practical ability to continue binding the company to new contracts and moving funds until the chop was physically recovered and the legal representative registration formally changed at the local Administration for Market Regulation office — a process that, based on comparable disputes in the sector, can take weeks to contest if the individual does not cooperate.

A portion of the company's compute infrastructure was financed through a provincial government AI development fund, structured as a low-interest facility rather than equity. This didn't appear as a red flag in the financials, since it was booked as ordinary debt financing rather than disclosed as a distinct government relationship in the investor-facing data room summary, and it meant a share of the burn rate was tied to a relationship the foreign investor had no visibility into or leverage over.

Modeling the facility's actual terms surfaced a renewal clause tied to the company maintaining a minimum headcount of R&D staff physically based in the fund's home province for the life of the facility — a condition that would directly constrain a future decision to consolidate engineering operations or relocate closer to a customer base, without that constraint being visible anywhere in the standard financial statements a diligence team would normally review.

Public-facing generative AI services in China require a filing and security assessment process that was still being clarified at the time of the deal. The company's roadmap assumed a compliant path that had not yet been tested for its specific model category, an assumption the pitch materials did not flag as unresolved, presenting the filing as a routine, near-complete formality rather than a genuinely open regulatory question.

Cross-referencing the company's planned enterprise-pilot rollout timeline against the filing's realistic approval window revealed a scenario where the two flagship pilot customers' contracted go-live dates preceded the earliest plausible date the filing could reasonably clear, based on comparable approval timelines for similarly-scoped models in the preceding cohort. This gap, if unaddressed, risked either a contractual breach with the pilot customers or a decision to launch ahead of formal clearance — a choice with materially different risk profiles that the committee had not been asked to weigh in on before the structural review surfaced it.