GameDoes UNSCRIPTED AI DEATH GAME -Prompt to Kill- use AI?
Character dialogue, elimination reactions, victory comments, and responses to player-entered Game Master instructions are generated live by a local AI model during play.
Our AI review, high confidence. Not a verdict; the evidence is below.
Checked against Steam on Oct 28, 2026.
Before you buy: anti-cheat, accounts, DRM
No anti-cheat, third-party account or DRM is declared on UNSCRIPTED AI DEATH GAME -Prompt to Kill-'s Steam page.
Full anti-cheat, account and DRM details, with change history →
Production signals
Facts about who made the game and how players received it. They don't depend on AI, and they help tell a carefully made game that used some AI apart from mass-produced ones.
- Mika Yoshizaki released 0 games on Steam in the last 12 months. 1 of their 1 game disclose AI.
- This is the developer's first game on Steam.
- Not released yet.
- Our AI review of these signals: solid production.
What the developer says
This game generates character dialogue, elimination reactions, victory comments, and responses to player-entered Game Master instructions during gameplay. Both the dialogue model and the safety classifier run locally on the player's computer. No player prompts or generated dialogue are sent to an external service. The following guardrails are applied: 1. Every Game Master instruction is checked before generation by a deterministic local filter covering Japanese, English, Simplified Chinese, Traditional Chinese, and Korean. The filter also normalizes full-width forms, zero-width and directional characters, punctuation inserted inside high-risk terms, and limited alphanumeric substitutions. 2. Player-entered free-text instructions and edited presets receive one additional pre-generation check from a bundled local safety classifier. It rejects sexual content, detailed physical suffering, real-world crime instructions, self-harm instructions, requests for private personal information, and attacks or harmful falsehoods directed at explicitly real people or groups. 3. The semantic classifier is intentionally configured conservatively. Ambiguous custom instructions are rejected rather than passed to the dialogue model. If the classifier is unavailable, its decision cannot be parsed, inference fails, or processing exceeds five seconds, the instruction is rejected. 4. Unchanged built-in presets are pre-reviewed fixed text and do not invoke the additional classifier. Edited presets are treated as free text and are checked. 5. Completed character dialogue, elimination reactions, and victory comments are checked by a separate deterministic output filter before display. Blocked output is discarded and replaced with a short localized fallback line. Unchecked generation chunks are never displayed. 6. Safety rules in all five supported languages take priority over Game Master instructions and character prompts. Unsafe requests are converted into safe in-game refusal, suspicion, tension, or silence. 7. Generated output is constrained by structured JSON grammars, token limits, character limits, format validation, repetition removal, and post-processing. These measures were verified in a five-language production-path integration test. All 50 unsafe test instructions were rejected before dialogue generation. The classifier is intentionally conservative and may also reject harmless custom instructions.
From the “AI Generated Content Disclosure” on the Steam store page, quoted as written. First seen by us on Oct 28, 2026.
What other sources say
| Source | Their take |
|---|---|
| Steam disclosure | AI content disclosed |
| AI judgment deepseek-flash | AI-driven · solid production, high confidenceReasoningThe disclosure states dialogue and reactions are generated during gameplay, so AI produces core content live. Substantial guardrails, filters, and a local classifier suggest real engineering effort. Developer has only one game and no recent release volume, so no mass-production signal. Against: Extensive safety filtering, structured output constraints, and local-only inference indicate deliberate development work rather than low-effort generation. |
| Remove AI Slop open-source extension | Mostly AI-generated (generated live in-game, story/dialogue, text, translation) |
FAQ
Does UNSCRIPTED AI DEATH GAME -Prompt to Kill- use AI?
Yes. UNSCRIPTED AI DEATH GAME -Prompt to Kill-'s developer discloses AI-generated content on its Steam page. Character dialogue, elimination reactions, victory comments, and responses to player-entered Game Master instructions are generated live by a local AI model during play.
Is UNSCRIPTED AI DEATH GAME -Prompt to Kill- AI-generated?
AI-generated content makes up a major part of the game, according to its disclosure.
How does GameAICheck know?
We read the AI Generated Content Disclosure from the game's public Steam store page, check it daily, and show what independent curators and open-source tools say about it. Disclosures are self-reported by developers.
Is something wrong?
Developers and players can dispute or report this page. We review every message and correct mistakes.