SoftPrick Registers Two Patents for AI Privacy Diagnostics and Agent Control
IT DAILY ·
✦ AI Summary
SoftPrick said on the 12th that it has registered two patents.
The registered patents are "Multimodal Attack Method for Diagnosing Privacy Vulnerabilities in Artificial Intelligence Models" and "Security Control Method for Prompts and System Behaviors Executed by an AI Gateway."
The former is a technology that uses multiple input formats to pre-diagnose the possibility of sensitive information exposure in AI models, while the latter is a technology that controls prompts and system behaviors between AI and enterprise systems in corporate environments based on security policies.
SoftPrick said on the 12th that it has registered two patents. The newly registered patents were developed to respond to security threats arising from the adoption of generative AI and AI agents. The patents cover multimodal AI model privacy vulnerability diagnostic technology and AI gateway security control technology.
The registered patents are "Multimodal Attack Method for Diagnosing Privacy Vulnerabilities in Artificial Intelligence Models" and "Security Control Method for Prompts and System Behaviors Executed by an AI Gateway."
Of the two, "Multimodal Attack Method for Diagnosing Privacy Vulnerabilities in Artificial Intelligence Models" is one of five AI privacy vulnerability diagnostic patents SoftPrick filed last year.
For multimodal AI, it is difficult to identify vulnerabilities with text-based attacks alone. This technology is designed to pre-diagnose the possibility of sensitive information exposure in AI models through attack scenarios that use multiple input formats.
Through the technology titled "Security Control Method for Prompts and System Behaviors Executed by an AI Gateway," SoftPrick is working on technology that controls prompts and system behaviors between AI and enterprise systems in corporate environments based on security policies. The technology can control and track prompts, access rights, and behaviors related to AI agents.
SoftPrick is developing technologies for LLM gateways and AI gateways based on its accumulated capabilities in certification, authorization, access control, and policy enforcement technologies in the API management and security field. It is also pursuing a technology that uses LLMs to analyze security guidelines and automatically generate vulnerability inspection rules and diagnostic scripts, and is also pursuing a plan to link this with its vulnerability diagnostic solution F-Avata.
Commenting on the background to this trend, Kim Dong-cheol, CEO of SoftPrick, said that as attackers' use of AI advances vulnerability discovery and attack automation, human-centered responses alone face limits in terms of speed and scale. He added that the defense side also needs to use AI, and that companies need to verify the safety of the AI they use and control access and behavior.
Source: IT DAILY · Kim Ho-jun
Original: https://www.itdaily.kr/news/articleView.html?idxno=242111
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.