Responsible Intelligence / Responsible AI Policy
AI can assist, but automation is not authority and consequential systems need evidence, limits, and accountable ownership.
Public draft · Version 2026-09-20
A public draft commitment. AI output, including a fluent answer, is not proof that an action occurred or a fact is true.
01Our position
DRKNYT uses artificial intelligence as a tool for assistance, exploration, analysis, creation, and operation. AIArtificial intelligence: software that produces or interprets material from patterns in data.Open the source can extend human capability, but fluency is not truth, automation is not authority, and a model response is not proof that an action occurred.
We remain responsible for the systems we choose, the boundaries we set, the data we provide, the actions we permit, and the claims we publish.
02Commitments
- Consequential AIArtificial intelligence: software that produces or interprets material from patterns in data.Open the source -enabled systems should have an identifiable human or organizational owner, and human review must be meaningful rather than ceremonial.
- AIArtificial intelligence: software that produces or interprets material from patterns in data.Open the source output alone is not authoritative proof of identity, ownership, payment, consent, legal status, safety, eligibility, or completion of an external action.
- We seek to provide models only the information needed for the task and keep private data private even when a model processes it.
- We distinguish generated content, inference, user assertion, sourced evidence, and authoritative external results rather than erasing uncertainty.
- High-impact actions should be separated from suggestion and protected by validation, permission, limits, and observable confirmation.
03Fairness, evaluation, and improvement
Models can reproduce exclusions and stereotypes present in their training data and in the systems around them. We test important outputs and workflows for unequal treatment, representation, accessibilityMaking information and controls usable by people with different abilities and ways of interacting.Open the source barriers, and concentration of error or burden.
No AIArtificial intelligence: software that produces or interprets material from patterns in data.Open the source system is neutral, infallible, or complete. We commit to bounded choices, consequential decision-making that can be reviewed, correction of verified harm, and policy revision as our understanding and systems develop.
Questions or a barrier?
Contact DRKNYT with the affected page and what you were trying to do. Share only the personal information needed to explain the issue.