AI Innovation for Public Safety: Castlepoint's Success with Hatch Dimension X, Singapore
Singapore’s information security environment
Singapore’s strong regulatory and cybersecurity environment makes it a global leader in regulatory technology (RegTech) and national security innovation. With increasing data volume and exposure to cyber threats, local organisations must manage sensitive information with precision and transparency. Castlepoint’s explainable AI approach aligns with Singapore’s frameworks for ethical and responsible AI, reinforcing resilience across critical sectors.
Local organisations face threats from foreign state actors due to their large data systems and interconnected networks. To address these risks, Singapore has developed national strategies for cybersecurity and AI, along with broader frameworks like Smart Nation 2.0 and the Digital Enterprise Blueprint. These initiatives emphasise resilience, cyber safety, and ethical AI use. International collaboration is also a key focus, and Castlepoint’s autoclassification technology is already being used by some of Singapore’s key allies, Australia, New Zealand, and the UK.
Our Mission: Explainable AI for Sensitive Data Classification
Our project focused on applying our Explainable Artificial Intelligence (XAI) capability to automate the classification of sensitive data based on its content and context, in a Singapore national security framework. The goal was to enhance national security and data governance by:
- Extracting current sensitivity and security labels from document headers/footers
- Suggesting appropriate target security markings for unstructured documents
- Providing clear, actionable reasoning for classification decisions
- Recommending down-classification strategies where appropriate
- Auditing over-, under-, and unclassified records
- Integrating with Microsoft Purview for label governance.
We used an international national security taxonomy as the basis for a classification logic model, mapped to the Singapore government context. We consulted closely with the project team to determine the most appropriate markings to apply, based on Singaporean government policy and practice.
Key Achievements
Working together, we were able to achieve the core targets: deploying the relevant ontologies; matching content accurately; iterating the matching; and showing recommendations for labels in a Singapore national security context.
We then had scope to achieve significant stretch targets with the solution, including:
- Introduction of GenAI as an option into the matching pipeline, additional to the CUI ontology matching
- Development of regular expression (regex) mapping for sensitivity labelling, specific to the Singapore national security and privacy context
- Introduction of a secure, air-gapped Small Language Model (SLM) to provide additional commentary on label recommendations
- Inclusion of our Purview capability, to push recommended labels to M365.
These features were tested using synthetic and publicly available data, curated in collaboration with HTX. The solution was deployed in a secure cloud environment, allowing both teams to iterate and validate the classification models effectively.
Collaboration and Impact
Castlepoint worked closely with HTX through workshops, demos, and live testing, achieving classification results in seconds. This collaboration validated explainable AI as a practical tool for secure data management and set the stage for future applications across government, defence, and critical infrastructure.
What's Next
Castlepoint’s participation in Hatch Dimension X has demonstrated the power of Explainable AI to enhance public safety and regulatory compliance in Singapore’s national security context. We’re excited to continue building on this success, and we look forward to future opportunities to support government agencies with transparent, scalable, and secure data governance solutions.
We extend our sincere thanks to HTX, Hatch, and the Ministry of Home Affairs for their partnership and support throughout the journey.
Learn more about our Expainable AI Capability and enhance your organisations governance and security. Get in touch today.
Originally published at castlepoint.systems.