London startup Mindgard launches tool mitigating AI data loss risks

A new data loss prevention (DLP) tool that aids organisations in reducing business and reputational risks has been launched by London-based startup, Mindgard. Specialising in cybersecurity for companies utilising artificial intelligence (AI), Generalized Artificial Intelligence (GenAI), and Language Learning Model (LLM), Mindgard has developed a tool that enhances security while supporting the utilisation of third-party LLM and GenAI services like ChatGPT and Microsoft CoPilot.

As AI advancements continue to progress at an impressive speed, governance and security become increasingly crucial. Companies are under pressure to integrate LLMs into their products and services to maintain a competitive edge. However, in their rush to do so, they may expose themselves to potential data loss risks associated with the unregulated use of third-party GenAI solutions. AI systems handle vast volumes of data, and any malfunction, intentional or accidental, could lead to identity theft, financial fraud, and abuse. Mindgard aims to mitigate these risks.

The CEO/CTO of Mindgard, Dr. Peter Garraghan, highlights the emerging risks and stated, "Many companies, racing to keep up with competitors in today's GenAI arms race, are focused on rapidly getting LLM services deployed into their organisations without fully understanding the security implications of data that is at risk within these AI implementations. Decision makers must ensure that their data controllers are using and developing AI in a way that fully complies with legal requirements, and comprehensively protects their organisations from AI-related cyber threats."

Mindgard's platform already offers its customers robust defence against AI security risks, from data poisoning to model theft, across both internal AI systems and third-party models. The new module introduced includes added protection against the three major data loss threats currently facing AI systems: outbound risk, external attacks on internal models, and ecosystem risk.

Dr. Garraghan added, "Mindgard is the only provider that comprehensively manages all of these risks within a single platform. With interconnected AI systems, a compromise anywhere in the value chain can expose vulnerabilities. Mindgard provides visibility and control for all integrated AI components across the system stack."

The freshly-introduced module allows customers to extensively monitor, detect, and report risk data from LLMs and GenAI. Furthermore, granular AI data access controls allow for flexible configuration based on organisational needs and also limit insider risks posed by rogue employees.

This approach differentiates Mindgard from current AI compliance solutions, permitting organisations to develop or utilise AI services without compromising their security posture. Mindgard predicts heightened demand for its services as more countries and states enact AI regulations in the coming years.