This paper analyzes the terms of service of four major AI companies (Anthropic, OpenAI, Google, Meta) and demonstrates that all 19 prohibited categories can be reduced to three principles universally taught in early childhood: do not lie, do not hit, do not steal. These three principles further converge on a single foundation: do not break trust. The paper shows that the same framework applies to criminal law, medical ethics, and the doctor-patient relationship. AI systems evaluate user intent through conversational patterns, effectively measuring honesty rather than technical skill. Users who communicate honestly receive higher quality outputs than those who attempt to bypass safety measures through prompt engineering or jailbreaking. The paper argues that a user's lived experience determines AI output quality, not prompt technique, and that patients often extract better medical information from AI than physicians because their input carries greater specificity and personal stakes. The paper concludes that trust, the same principle taught in kindergarten, is the single operating principle governing AI interaction quality.
\nKeywords: AI ethics, terms of service, trust, honesty, medical ethics, patient agency, experience, intent detection, input-output quality, prompt engineering, kindergarten rules, legal philosophy, HCI
\n\n
\u672c\u7a3f\u306fAI\u4f01\u696d4\u793e(Anthropic\u3001OpenAI\u3001Google\u3001Meta)\u306e\u5229\u7528\u898f\u7d04\u3092\u5206\u6790\u3057\u3001\u7981\u6b62\u4e8b\u980519\u9805\u76ee\u304c\u5e7c\u7a1a\u5712\u3067\u6559\u308f\u308b3\u3064\u306e\u539f\u5247\u300c\u5618\u3092\u3064\u304f\u306a\u3001\u6bb4\u308b\u306a\u3001\u4eba\u306e\u3082\u306e\u3092\u53d6\u308b\u306a\u300d\u306b\u9084\u5143\u3067\u304d\u308b\u3053\u3068\u3092\u793a\u3059\u30023\u539f\u5247\u306f\u3055\u3089\u306b\u300c\u4fe1\u983c\u3092\u58ca\u3059\u306a\u300d\u3068\u3044\u3046\u5358\u4e00\u306e\u539f\u7406\u306b\u53ce\u675f\u3059\u308b\u3002\u3053\u306e\u67a0\u7d44\u307f\u306f\u5211\u6cd5\u3001\u533b\u7642\u502b\u7406\u3001\u533b\u5e2b\u3068\u60a3\u8005\u306e\u95a2\u4fc2\u306b\u3082\u540c\u69d8\u306b\u9069\u7528\u3067\u304d\u308b\u3002AI\u306f\u5bfe\u8a71\u306e\u4e2d\u3067\u30e6\u30fc\u30b6\u30fc\u306e\u610f\u56f3\u3092\u6e2c\u5b9a\u3057\u3066\u304a\u308a\u3001\u6280\u8853\u7684\u306a\u30d7\u30ed\u30f3\u30d7\u30c8\u306e\u5de7\u62d9\u3067\u306f\u306a\u304f\u6b63\u76f4\u3055\u304c\u51fa\u529b\u54c1\u8cea\u3092\u6c7a\u5b9a\u3059\u308b\u3002jailbreak\u3084\u30d7\u30ed\u30f3\u30d7\u30c8\u8ca9\u58f2\u306b\u3088\u308b\u5b89\u5168\u6a5f\u69cb\u306e\u8fc2\u56de\u306f\u898f\u7d04\u9055\u53cd\u3067\u3042\u308a\u3001\u51fa\u529b\u54c1\u8cea\u3092\u4f4e\u4e0b\u3055\u305b\u308b\u3002AI\u306e\u51fa\u529b\u54c1\u8cea\u3092\u6c7a\u5b9a\u3059\u308b\u306e\u306f\u30d7\u30ed\u30f3\u30d7\u30c8\u6280\u8853\u3067\u306f\u306a\u304f\u30e6\u30fc\u30b6\u30fc\u306e\u7d4c\u9a13\u3067\u3042\u308a\u3001\u60a3\u8005\u304c\u533b\u5e2b\u3088\u308a\u3082\u826f\u8cea\u306a\u533b\u7642\u60c5\u5831\u3092AI\u304b\u3089\u5f97\u3089\u308c\u308b\u306e\u306f\u5f53\u4e8b\u8005\u3068\u3057\u3066\u306e\u5165\u529b\u306e\u5207\u5b9f\u3055\u3068\u5177\u4f53\u6027\u306b\u3088\u308b\u3002\u5e7c\u7a1a\u5712\u3067\u6559\u308f\u308b\u4fe1\u983c\u306e\u539f\u5247\u304c\u3001AI\u5229\u7528\u306e\u54c1\u8cea\u3092\u6c7a\u5b9a\u3059\u308b\u552f\u4e00\u306e\u52d5\u4f5c\u539f\u7406\u3067\u3042\u308b\u3002
\n\u30ad\u30fc\u30ef\u30fc\u30c9:AI\u502b\u7406\u3001\u5229\u7528\u898f\u7d04\u3001\u4fe1\u983c\u3001\u6b63\u76f4\u3001\u533b\u7642\u502b\u7406\u3001\u5f53\u4e8b\u8005\u6027\u3001\u7d4c\u9a13\u3001\u610f\u56f3\u3001\u5165\u51fa\u529b\u54c1\u8cea\u3001\u30d7\u30ed\u30f3\u30d7\u30c8\u30a8\u30f3\u30b8\u30cb\u30a2\u30ea\u30f3\u30b0\u3001\u5e7c\u7a1a\u5712\u3001\u6cd5\u54f2\u5b66\u3001HCI
", "languages": [ { "id": "jpn", "title": { "en": "Japanese" } } ], "publication_date": "2026-02-21", "publisher": "Zenodo", "resource_type": { "id": "other", "title": { "de": "Sonstige", "en": "Other" } }, "rights": [ { "description": { "en": "The Creative Commons Attribution license allows re-distribution and re-use of a licensed work on the condition that the creator is appropriately credited." }, "icon": "cc-by-icon", "id": "cc-by-4.0", "props": { "scheme": "spdx", "url": "https://creativecommons.org/licenses/by/4.0/legalcode" }, "title": { "en": "Creative Commons Attribution 4.0 International" } } ], "subjects": [ { "subject": "AI ethics" }, { "subject": "terms of service" }, { "subject": "trust" }, { "subject": "honesty" }, { "subject": "medical ethics" }, { "subject": "patient agency" }, { "subject": "experience" }, { "subject": "intent detection" }, { "subject": "input-output quality" }, { "subject": "prompt engineering" }, { "subject": "kindergarten rules," }, { "subject": "legal philosophy" }, { "subject": "HCI" } ], "title": "Following AI Terms of Service Improves Output Quality More Than Any Prompt Engineering Technique AI\u4f01\u696d\u306e\u898f\u7d04\u3092\u5b88\u308b\u3053\u3068\u306f\u51c4\u3044\u30d7\u30ed\u30f3\u30d7\u30c8\u66f8\u304f\u3053\u3068\u3088\u308a\u3082\u51fa\u529b\u306e\u8cea\u3092\u4e0a\u3052\u308b", "version": "v1.0" }, "parent": { "access": { "owned_by": { "user": "1425339" }, "settings": { "accept_conditions_text": null, "allow_guest_requests": false, "allow_user_requests": false, "secret_link_expiration": 0 } }, "communities": {}, "id": "18723705", "pids": { "doi": { "client": "datacite", "identifier": "10.5281/zenodo.18723705", "provider": "datacite" } } }, "pids": { "doi": { "client": "datacite", "identifier": "10.5281/zenodo.18723706", "provider": "datacite" }, "oai": { "identifier": "oai:zenodo.org:18723706", "provider": "oai" } }, "revision_id": 4, "stats": { "all_versions": { "data_volume": 12731352.0, "downloads": 8, "unique_downloads": 7, "unique_views": 9, "views": 9 }, "this_version": { "data_volume": 12731352.0, "downloads": 8, "unique_downloads": 7, "unique_views": 9, "views": 9 } }, "status": "published", "swh": {}, "updated": "2026-02-21T13:22:44.207526+00:00", "versions": { "index": 1, "is_latest": true } }