Making AI Governance Part of Your Code, Not Just Paperwork
Abstract
AI governance often remains confined to policy documents, periodic reviews, and compliance checklists that may not be revisited after approval. AI systems, however, continue to change: models are retrained, datasets are replaced, prompts are revised, and system components are updated. Without continuous validation, the gap between approved governance requirements and the system’s current behavior may remain undetected until an audit, incident, or user complaint occurs.
This talk presents a practical approach for closing that gap by translating governance requirements into automated and testable engineering controls. Requirements related to privacy, fairness, security, transparency, data quality and accountability are validated alongside unit and system tests. For example, if a customer-service AI must not disclose personal information, representative inputs can be used to test whether its responses comply with that requirement. A failed check can prevent release until the issue is corrected or an exception is formally reviewed and approved.
When a governance requirement changes, its associated tests are updated, and the affected parts of the system are revalidated. Engineers receive clear and timely results, while compliance officers gain traceable evidence of what was tested, when it was tested, and how failures or exceptions were addressed. By connecting policy requirements directly with software testing, this approach transforms AI governance from a one-time documentation exercise into a continuous and verifiable part of AI development.
- Published in
- Proceedings of the Raptors Conference, Vol. 1 (2026)
- Conference
- Raptors Conference 2026, 21 October 2026, online
- Accepted
- 2026-10-12
- DOI
- https://doi.org/10.68495/rc.2026.C042
- Presentation
- Slides are not yet available for this paper.
- Licence
- © 2026Manish Gupta. Published under Creative Commons Attribution 4.0 International (CC BY 4.0).
- Publisher
- Hackathon Raptors, Community Interest Company 15557917, London, United Kingdom
- Permanent address
- https://proceedings.raptors.dev/gupta-making-ai-governance-part
Open access. Every entry in the Proceedings is free to read, with no charge to authors and no charge to readers. Slides are served from this site, so they stay available at the paper's permanent address.
Cite this paper
Manish Gupta (2026). Making AI Governance Part of Your Code, Not Just Paperwork. Proceedings of the Raptors Conference, Vol. 1 (2026). Hackathon Raptors, London. https://doi.org/10.68495/rc.2026.C042
@inproceedings{raptors2026-c042,
title = {{Making AI Governance Part of Your Code, Not Just Paperwork}},
author = {Gupta, Manish},
booktitle = {Proceedings of the Raptors Conference},
volume = {1},
year = {2026},
publisher = {Hackathon Raptors},
address = {London, United Kingdom},
doi = {10.68495/rc.2026.C042},
url = {https://doi.org/10.68495/rc.2026.C042},
abstract = {AI governance often remains confined to policy documents, periodic reviews, and compliance checklists that may not be revisited after approval. AI systems, however, continue to change: models are retrained, datasets are replaced, prompts are revised, and system components are updated. Without continuous validation, the gap between approved governance requirements and the system’s current behavior may remain undetected until an audit, incident, or user complaint occurs. This talk presents a practical approach for closing that gap by translating governance requirements into automated and testable engineering controls. Requirements related to privacy, fairness, security, transparency, data quality and accountability are validated alongside unit and system tests. For example, if a customer-service AI must not disclose personal information, representative inputs can be used to test whether its responses comply with that requirement. A failed check can prevent release until the issue is corrected or an exception is formally reviewed and approved. When a governance requirement changes, its associated tests are updated, and the affected parts of the system are revalidated. Engineers receive clear and timely results, while compliance officers gain traceable evidence of what was tested, when it was tested, and how failures or exceptions were addressed. By connecting policy requirements directly with software testing, this approach transforms AI governance from a one-time documentation exercise into a continuous and verifiable part of AI development.}
}
TY - CONF TI - Making AI Governance Part of Your Code, Not Just Paperwork AU - Gupta, Manish T2 - Proceedings of the Raptors Conference VL - 1 PY - 2026 DA - 2026/10/12 PB - Hackathon Raptors CY - London, United Kingdom DO - 10.68495/rc.2026.C042 UR - https://doi.org/10.68495/rc.2026.C042 AB - AI governance often remains confined to policy documents, periodic reviews, and compliance checklists that may not be revisited after approval. AI systems, however, continue to change: models are retrained, datasets are replaced, prompts are revised, and system components are updated. Without continuous validation, the gap between approved governance requirements and the system’s current behavior may remain undetected until an audit, incident, or user complaint occurs. This talk presents a practical approach for closing that gap by translating governance requirements into automated and testable engineering controls. Requirements related to privacy, fairness, security, transparency, data quality and accountability are validated alongside unit and system tests. For example, if a customer-service AI must not disclose personal information, representative inputs can be used to test whether its responses comply with that requirement. A failed check can prevent release until the issue is corrected or an exception is formally reviewed and approved. When a governance requirement changes, its associated tests are updated, and the affected parts of the system are revalidated. Engineers receive clear and timely results, while compliance officers gain traceable evidence of what was tested, when it was tested, and how failures or exceptions were addressed. By connecting policy requirements directly with software testing, this approach transforms AI governance from a one-time documentation exercise into a continuous and verifiable part of AI development. LA - en ER -