Responsible AI Policy

Globenova builds infrastructure for the accountable use of artificial intelligence in sensitive environments. We hold our own work to the principles we expect of the systems we are built to govern.

01Why we publish this

Our products exist so that organisations can adopt AI without losing accountability for what it does. A company in that position should be willing to state its own commitments plainly. This policy does that. We would rather state a limit clearly than overstate a capability.

02Human accountability and oversight

We design our systems so that AI supports human decision-makers and does not replace human judgment or accountability. A named, authorised person remains answerable for a consequential decision. We do not build systems whose purpose is to remove a human from such a decision.

03What we do not build

We do not develop autonomous weapons or systems that select or engage targets without meaningful human control. We do not build tooling intended for unlawful surveillance. Our platform inspects and governs how AI is used; it does not generate, direct, or apply force.

04Transparency and auditability

Our platform is designed to record its decisions to a tamper-evident log that an authorised reviewer can examine after the fact. We design for explanation rather than opacity, so that an organisation can show what happened and why.

05Reliability and fail-closed safety

When our systems are uncertain or impaired, they are designed to fail closed, blocking or escalating rather than passing content through unchecked. Safety is the default behaviour, not an optional setting.

06Lawfulness

We design our products to operate within applicable law. Depending on the deployment, that includes privacy law, export and controlled-goods rules, and, where relevant to a customer's mission, international humanitarian law and the rules of the jurisdictions they serve.

07Privacy and data minimisation

We minimise the personal data we and our systems handle, keep sensitive data in the customer's custody, and provide synthesis capabilities so that models can be developed and tested without exposing protected source data.

08Fairness

We treat classification accuracy as a safety property. Unexplained disparities in how our systems treat comparable inputs are defects to be investigated and corrected, not outcomes to be accepted.

09Security by design

Security and sovereignty are designed into our products from the outset rather than added afterwards. We welcome scrutiny of our security through our Responsible Disclosure Policy.

10Sovereignty

Our customers keep custody and control of their data and their decisions. We do not require sensitive data to leave the customer's jurisdiction in order for our products to work.

11Alignment with recognised frameworks

Our approach draws on widely recognised guidance, including the OECD AI Principles, Canada's Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI, the Government of Canada Directive on Automated Decision-Making, the NIST AI Risk Management Framework, and allied principles for the responsible use of AI in defence. We align our design to these frameworks; references to them describe the standards we work toward, not a claim of certification by any body.

12Governance and review

We review this policy as our products and the wider field evolve. Questions or concerns can be sent to info@globenova.ai.