Redefining Ai Privacy: Synology’s Commitment to Secure and Private Cloud Solutions

Earlier this year, a tech giant introduced an AI feature that takes screen snapshots and stores them locally on its upcoming AI PCs for future recall. Despite the company’s assurances of encryption, a spokesperson from the Office of the Australian Information Commissioner (OAIC) emphasised the serious privacy risks tied to AI systems that process personal information, urging organisations to perform privacy impact assessments on such technologies. In addition, the OAIC released new guidance on privacy and the development of generative AI models, setting clear regulations for organizations on the collection and use of customer data.

This incident highlights the broader privacy concerns surrounding AI-powered technologies. For instance, public cloud photo management services often use facial recognition to organise photos, offering convenience but raising privacy issues. While companies emphasise encryption and user control, privacy advocates consistently express concerns about data collection, potential misuse, and the creation of facial recognition databases.

As a private cloud solution provider, Synology prioritises data security and privacy in our product design. This commitment stems from our understanding of the sensitive nature of customer data entrusted to our systems. While we’ve integrated AI functionalities into our ecosystems and support systems, we ensure that privacy rights are embedded in our design principles, taking precedence over functional efficiency.

Ensuring Data Privacy with Pre-Trained AI Models and Private Cloud Architecture

Synology integrates AI facial and object recognition techniques into its photo management application, Photos, to enhance photo management and search experiences. The AI recognition models used in Photos are pre-trained with publicly available datasets during development, prior to any user interactions with the application.

Unlike photo management SaaS which stores users’ data in the solution provider’s data centers, all Synology users’ data resides on their private devices, ensuring full data ownership. This means Synology does not and can never use users’ data for AI model training.

Similarly, Synology’s surveillance solution incorporates advanced AI features like people and vehicle detection, facial recognition, and object tracking, enhancing security by reducing false alarms. These pre-trained AI models operate effectively within the user’s network. This approach, combined with data encryption and access control settings, offers powerful monitoring tools while ensuring user privacy and data security, eliminating the risk of reliance on third-party cloud services.

Addressing Shadow AI and Data Leaks with a Dedicated Admin Console

Synology also integrates large language models (LLMs) into our Synology Office Suite, allowing users to incorporate generative AI as a writing assistant for emails, slides, documents, and spreadsheets.

To help organisations maintain privacy and mitigate the risk of shadow AI while using generative AI tools, we will introduce an AI admin console for system administrators in 2025 Q1. This tool provides granular control over the AI models and versions that can be deployed within the organisation, permissions for models allowed across organisations, groups, or individuals, and de-identification parameters including regional-specific information such as Social Security Number (SSN). This allows businesses to leverage AI tools safely without compromising sensitive data.

Related: Managed IT services Sydney

Additionally, we embedded a proprietary de-identification mechanism in the console to help customers protect their organisations’ sensitive information. For instance, when a user generates a prompt using the AI writing assistant that includes sensitive information such as IP addresses and emails, the system first performs de-identification, masking sensitive data with generalised labels such as <ip_address> and <email>. Once the generative AI receives the sanitised prompt and produces a response with generalised labels, the system reintegrates the original information back into the response within the Synology system. This ensures that sensitive information never leaves the user’s network while using generative AI tools.

Robust AI Customer Support Through GDPR-Compliant De-Identification Processes

Synology also uses AI to enhance customer service experiences. We have built an AI support system that retrieves data from a knowledge database consisting of anonymised past technical support data.

To ensure our customers’ privacy is protected, the past support data used to build the database undergoes a comprehensive de-identification process. This includes methodologies such as Regex, Named Entity Recognition (NER), checksum validation, and context analysis to detect and anonymise sensitive information. Additionally, it employs query guardrails to prevent sharing sensitive information or inadvertently instructing users to delete critical data.

When a new support ticket is submitted, our support system would perform the same de-identification and guardrail processes before retrieving insights and similar feedback from the support knowledge database. Most importantly, the building of the database as well as processing new support tickets both go through a GDPR-compliant environment to protect our customers’ privacy.

Synology’s Ongoing Commitment to Data Privacy

Amidst these developments, Synology remains committed to prioritising privacy. With our privacy-first design principles and data safety guidelines, users can trust that their data is secure and protected, allowing them to benefit from AI advancements without compromising their privacy.

Steven Liang is Manager of Generative AI Application at Synology.

Last Updated on July 10, 2026 by Steven Liang