AI-Assisted Development Policy
We ask buyers to hold our product to a standard about what AI may and may not decide. It would be inconsistent not to state the same thing about how we build it.
Where AI is permitted
Alethia Software permits AI-assisted software development as an engineering productivity tool. AI tools may assist with scaffolding, refactoring, tests, documentation and implementation suggestions.
Human review is mandatory
All production code must be reviewed, understood, tested and accepted by Alethia engineers before merge. AI output is not accepted directly into production without human engineering review.
Developers remain responsible for correctness, security, maintainability, license compliance and alignment with product requirements. That responsibility is not transferable to a tool.
Areas requiring heightened review
AI-assisted changes to any of the following require additional human scrutiny before merge:
- Architecture
- Security
- Authentication and authorization
- Payment handling
- Compliance logic
- Export-controlled functionality
- Customer-sensitive logic
Evidence of human authorship
We maintain source control, pull-request review history, issue tracking, test results and release records as evidence of human authorship, engineering control and quality assurance. If a customer or an assessor needs to establish that a change was made under human engineering control, the record supports that.
How this relates to the product
The same principle governs CertiField itself: AI proposes, evidence supports, authorized practitioners decide, and the record preserves who decided what. See human-governed AI for how that is enforced in the product rather than merely stated.
Questions
Contact info@certifield.software for our software development documentation or to discuss this policy as part of a vendor assessment.