LLM Object Access Layer for Domain-Specific Privacy Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing large language models (LLMs) face challenges in managing dynamic data access control and privacy commitments across multiple domains, requiring continuous adjustments and configurations to minimize errors and ensure compliance with evolving privacy policies.

Innovation Solution

Implementing an LLM object access layer that determines user access levels and privacy demands for application objects, allowing configurable and extensible data access control, with logging capabilities to optimize access settings and reduce maintenance efforts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If LLMs are configured with domain specific settings and privacy commitments, then output accuracy and privacy compliance are improved, but system complexity and maintenance efforts increase

Engineering Contradiction:
Improveprivacy complianceVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent introduces an LLM object access layer as an intermediary component between the LLM and the application objects. This layer manages access control, filtering, and privacy commitments centrally, rather than embedding complex configuration logic throughout the system. The access layer acts as a mediator that handles privacy compliance and access control requirements, reducing overall system complexity while maintaining reliability.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the system into distinct layers: the LLM object access layer, the LLM itself, and the application objects. This segmentation allows each component to have a specific responsibility - the access layer handles privacy and access control, while the LLM focuses on generating responses. This modular structure reduces system complexity by separating concerns and making the system more manageable.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If LLMs are made configurable and extensible by customers, then adaptability to different domains is improved, but difficulty in managing access control increases

Engineering Contradiction:
Improvedomain adaptabilityVSAvoidaccess control management
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The LLM object access layer is designed as a universal component that can serve multiple domains and purposes. It provides a standardized interface for access control and filtering that works across different application objects and domains. This universal approach allows customers to extend the system to new domains without creating custom access control logic for each case, reducing the difficulty of managing access control while maintaining high adaptability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Manufacturing precision

If continuous adjustments are made to LLM configurations, then output accuracy is improved, but maintenance time and storage costs increase

Engineering Contradiction:
Improveoutput accuracyVSAvoidmaintenance time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent implements preliminary action by pre-configuring the LLM object access layer with filtering rules, access control policies, and domain-specific settings. These configurations are established in advance and can be reused across multiple scenarios. When new domains or requirements arise, the system can extend existing configurations rather than creating everything from scratch, significantly reducing maintenance time while maintaining output accuracy.

Inventive Principle:
Principle #10Preliminary action

4Loss of information

If detailed access control logging is implemented, then insights into access statistics are improved, but storage requirements increase

Engineering Contradiction:
Improveaccess statisticsVSAvoidstorage requirements
Core Design Contradiction:
Loss of informationVSQuantity of substance

Solution Approach 1:

The patent extracts the logging function as a separate capability within the LLM object access layer. Rather than storing all possible access data, the system selectively logs specific access statistics and metrics that provide valuable insights. This extraction approach allows the system to maintain detailed enough logging for analytical purposes while avoiding unnecessary storage of redundant or low-value data, balancing information quality with storage requirements.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentUS12541616B2Data access control for domain specific large language model services
Publication Date: 2026.02.03 SAP SE
  • US12541616B2 patent drawing
  • US12541616B2 patent drawing
  • US12541616B2 patent drawing

AI summary

The disclosure generally describes methods, software, and systems for to data access control for applications using large language models (LLM). A request to access attributes of an application object is received from a user. A LLM object access layer defining an access level of the user to each attribute of the application object is determined. The LLM object access layer is queried to determine accessible attributes of the attributes of the application object. A response is provided based on the accessible attributes of the application object to the user.