(PATENT) The Complying Language Model™ aka. "CLM™" - Secure, Compliant, and Trustworthy Language, Generative and Reasoning AI (2025)
Year: 2025
Type: article
Abstract: This report introduces the Complying Language Model (CLM), a novel architecture designed to address the inherent security and compliance failures of current Large Language Models (LLMs). The CLM aims to provide a secure, compliant, and trustworthy generative AI solution. "The Complying Language Model (CLM) Secure, Compliant, and Trustworthy Generative and Reasoning AI" highlights the significant vulnerabilities of existing LLMs, including prompt injection, data and model poisoning, and model inversion attacks, which can lead to sensitive data exposure and financial risks. It notes that the economic consequences of insecure AI are projected to reach $10.5 trillion annually by 2025, with data breaches costing U.S. organizations an average of over $10 million per incident. Furthermore, current LLM architectures are incompatible with major data protection regulations like GDPR and HIPAA, leading to substantial regulatory fines.
Author: Alpasan, Lois-Kleinner
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7078841596
FWCI:
Cites:
Cited by:
Related to: