복원(로컬 복원이라고도 함)은 AI가 생성한 결과를 원래의 기업 맥락으로 되돌리는 과정입니다. AI가 캡슐화된 데이터를 처리한 후, 원본 값과 보호된 값 사이의 로컬 저장 매핑이 AI 출력에 적용되어 캡슐화된 표현을 실제 기업 데이터로 대체합니다. 이 자동화된 과정을 통해 AI 결과는 로컬에서 복원됩니다.
Explanation
- When LLM Capsule encapsulates a document, it creates a local mapping table. The table records the relationship between each original sensitive value and its replacement. This mapping never leaves the enterprise environment.
- After AI processing completes, the restoration engine applies this mapping in reverse. Encapsulated tokens are replaced with original values. The result is an AI output that contains real names, real figures, and real references, ready for enterprise use.
- Restoration is deterministic. Every encapsulated value maps to exactly one original value. There is no ambiguity, no approximation, and no manual intervention required.
Example
AI processes an encapsulated contract and generates a summary: “The agreement between ENTITY_A and ENTITY_B provides for a payment of AMOUNT_1 due on DATE_1.”
Local restoration applies the mapping and produces: “The agreement between Acme Corp and GlobalTech provides for a payment of $2.4M due on March 15, 2026.”