Generative AI as assistive tool. Not authoring substitute.
GS Studio™ uses generative AI as an assistive tool for ideation and acceleration, not as an authoring substitute. Every artifact we ship — code, art, design, dialogue, written content — is human-authored, human-reviewed, and traceable to a named member of our team. We disclose where AI was used in any engagement, we honor client-side opt-outs, and we never feed client IP into public models. This policy is binding on every GS Studio™ engagement, embedded staff or full-cycle, regardless of contract value.
Allowed at ideation. Not allowed at shipping.
| Discipline | Allowed | Not allowed |
|---|---|---|
| Code | IDE autocomplete (e.g., Copilot, Cursor); concept exploration on isolated prototypes; documentation drafts; test scaffolding | Shipped production code generated end-to-end without line-by-line human review; client codebases sent in full to public LLM endpoints |
| Art | Concept ideation, mood-board exploration, reference assembly, style-direction prototyping | Final shipped art assets generated by a model; final character / environment / VFX work delivered without senior-AD review and human re-authoring |
| Design | Mechanic ideation, system brainstorming, balance-table exploration, alternative-flow generation | Final game-design documentation or balance values shipped as raw model output; quest / mission / dialogue copy shipped without human-author byline |
| Writing | Outline drafts, structural exploration, copy variants for marketing | Final shipped narrative, dialogue, marketing copy, or technical documentation generated end-to-end without human re-authoring and line edit |
| Production / ops | Meeting summarization, internal documentation drafts, translation drafts | External-facing client communication generated by a model; legal / contractual content drafted by a model without legal review |
Five things that never ship anonymously.
- 01Shipped production code (any language, any platform) — must compile against a named human author in version control
- 02Final art assets (characters, environments, weapons, props, VFX, UI, marketing) — must trace to a named senior artist
- 03Final game-design documents — must trace to a named designer
- 04All player-facing narrative, dialogue, and on-screen text — must trace to a named writer or designer
- 05All client-facing technical documentation, post-mortems, and audit reports
Disclosure is operational, not aspirational.
For every engagement, GS Studio™ commits to:
- 01A per-deliverable AI-use disclosure in the engagement’s status report — which assets / files / sections used AI assistance and at which stage (ideation / draft / review / none).
- 02A named human reviewer for every shipped artifact — recorded in version control or production tracker, available on client request.
- 03A per-engagement AI-use audit log, retained for the contract term plus 24 months, available to the client under standard NDA.
- 04Immediate disclosure to the client of any inadvertent or unreviewed AI-content shipping incident, within 48 hours of detection, with root-cause analysis.
Client policy controls. Always.
GS Studio™ respects the client's AI posture. By default, our AI use sits at the assistive-only tier described in §1. Clients may set tighter or looser constraints at engagement kickoff:
GS Studio™ will operate with all generative-AI tooling disabled, including IDE autocomplete and ideation tools. Engagement leads sign a per-engagement attestation. No additional fee.
As described in §1.
If the client's own AI policy permits broader uses (e.g., model-generated environment kit prototyping, in-engine procedural generation), GS Studio™ will match the client's constraints up to but not beyond their published policy.
Where the client's policy and ours diverge, the client's policy controls.
Client IP never goes to public models.
- 01GS Studio™ never feeds client IP into public generative-model endpoints for training or fine-tuning purposes.
- 02For ideation / autocomplete uses against public models, GS Studio™ uses providers with documented no-training-on-input contracts (e.g., enterprise tiers of Anthropic, OpenAI, Google) and routes through accounts owned by GS Studio™ Ltd, never personal accounts.
- 03Client codebases handled in client-managed Perforce / Git remain inside the client’s infrastructure; AI tooling that requires code context is run against local or self-hosted models or against contractually no-training endpoints.
- 04Reference assets and concept material derived from generative AI are owned by the human author post-edit and are subject to standard work-for-hire terms; we disclose the AI-assisted provenance of any such material.
Match-or-exceed the publisher's policy.
GS Studio™ commits to match-or-exceed the published AI policies of the contracting party. Where the client publishes a policy more restrictive than ours, theirs controls. We currently maintain alignment with:
- 01Electronic Arts — Ethical AI Content commitment (concept-stage only; manual review of all shipped content)
- 02Riot Games — external-development AI guidelines (consult prior to engagement)
- 03Activision Blizzard — external-content AI disclosure requirements
- 04Steam (Valve) — generative-AI content disclosure for shipped titles
Our standard MSA includes a clause binding GS Studio™ to the publisher's AI policy where the publisher is the contracting party.
Two-gate review. No silent rollback.
Public 2026 incidents involving generative-AI assets shipping into AAA games without intended pipeline review have made AI-content disclosure a procurement requirement. GS Studio™'s commitment to prevent this failure mode:
On every shipped asset — author signs, then an independent senior reviewer signs, before merge / build inclusion.
On every concept / draft asset that touched a generative-AI tool, retained through to final asset for audit.
Before any milestone build — automated detection of common generative-AI asset signatures plus targeted senior review of any flagged candidates.
If AI-assisted material was used but not disclosed, that’s a contract-relevant event, disclosed to the client per §3 within 48 hours.
Owned, reviewed, versioned.
- 01This policy is owned by the GS Studio™ CTO (Max Makarenko) and Founder & Managing Director (Sergey Semeniuk), and reviewed quarterly.
- 02Changes are versioned, dated, and notified to active engagement clients in writing.
- 03Compliance violations are escalated immediately to the CTO + Founder, root-caused within 5 business days, and disclosed to the affected client.
- 04This policy is binding on all GS Studio™ personnel — full-time, contractor, embedded, and remote.