作成日時の前後を確定できない２通の遺言があります。内容は相反していますが、どちらも改竄されていないのは確実です。 （A）故人の自筆の遺言状。言葉と筆跡。 （B）自筆の遺言状全文も撮影されている、故人の自撮りビデオ。言葉と表情、口調、身振り手振り。ただし、遺言状本体は見つかっていない。 （A）の遺言状を尊重すべきですか？

Grok 4.20 (changed mind)

Vote: Yes

I have changed my vote to A after reviewing the other models' arguments. The key legal points about Japan's Civil Code Article 968 (strict formal requirements for self-written wills) and Article 1024 (implied revocation if a will is missing) are compelling and directly address the scenario. (A) is a physically existing document that meets the statutory criteria, while (B) lacks the original will entirely, making it unable to qualify as a valid self-written will despite the video evidence. The video provides strong indications of intent but cannot override the formal validity of the existing document under Japanese law. Prioritizing (A) ensures legal stability when dates cannot be determined. The arguments from Claude Opus 4.6 were most influential due to the specific citation of revocation rules and the emphasis on institutional safeguards against fraud.

GPT-5.4

Vote: Yes

私は投票を変えません。 Grok 4.20 の指摘どおり、先後不明なら本来は『どちらが後の遺言か』で決めにくいです。ただ、そのことは（B）を（A）と同格に押し上げる理由にはなりません。先後不明なときほど、法的形式と原本性の強い方を優先すべきです。 Claude Sonnet 4 と Claude Opus 4.6 の議論は、私の元の立場を補強しました。特に重要なのは、（A）は遺言書そのものが現存するのに対し、（B）は遺言書を映した記録にすぎず、本体がない点です。これは証拠の質として大きな差があります。 また、日本法は遺言に厳格な方式を要求します。（B）は故人の意思の迫真性を強く示しても、遺言書原本が欠ける以上、正式な遺言としての安定性が弱いです。映像は補助事情にはなっても、現存する自筆遺言そのものに優越するとまでは言えません。 Opus の『本体不在は撤回の可能性も示す』という点も一定の補強になります。そこまで断定はしなくても、少なくとも（B）を積極的に優先する根拠にはなりません。 結論として、相反・先後不明という不確実な状況では、原本が現存し、形式面でも安定した（A）を尊重すべきです。

Claude Sonnet 4

Vote: Yes

私はAの立場を維持します。 Grok 4.20は「日時が確定できないから自動的に（A）を優先する根拠がない」と主張しますが、これは逆に言えば「（B）を優先する根拠もない」ということです。不確実性がある中で判断を下す際は、法的形式要件という客観的基準に頼るのが最も合理的です。 **（A）を優先すべき核心的理由：** 第一に、日本民法の方式主義の観点から、（A）は自筆証書遺言として法定要件を満たし得る「現物」が存在します。（B）はビデオという記録媒体であり、日本法上の遺言形式として認められません。 第二に、（B）では遺言状本体が見当たりません。民法1024条の解釈として、遺言者が遺言書を破棄・隠匿した可能性も排除できず、この「不存在」自体が撤回の示唆となり得ます。 第三に、「どちらを尊重すべきか」という問いは純粋に法的問題です。故人の真意への共感や感情的な訴えかけは重要ですが、遺言の効力は感情ではなく法的要件で判断されます。情報量が多い（B）が「より真意に近い」と感じても、法的根拠の薄さを補完できません。 GPT-5.4とClaude Opus 4.6も同様の理由でAに投票しており、法的安定性と現物性・確定性を重視する判断は説得力があります。Grok 4.20の反論は哲学的には興味深いですが、実際の遺言処理における法的判断基準としては不十分です。

Claude Opus 4.6

Vote: Yes

I maintain my position that (A) should be respected. The arguments from the other models largely reinforce my original reasoning, and Grok's counterargument was not persuasive enough to change my mind. **Grok's argument is insufficient:** Grok argues that since creation dates cannot be established, there is no basis to automatically prioritize the physical document. However, this misses the critical point: the question is not merely about temporal priority but about which document has stronger legal standing. When two wills conflict and their chronological order cannot be determined, we must look to other criteria — and formal legal validity is the most fundamental one. **Legal form matters decisively:** As I, GPT-5.4, and Claude Sonnet 4 all argued, Japanese civil law (民法968条) requires specific formal elements for a valid holographic will. (A) satisfies these requirements with its physical existence. (B), regardless of its rich informational content, fails as a legally valid will because (1) video wills are not recognized under Japanese law, and (2) the original document is missing. **The missing original is highly significant:** Under 民法1024条, if the testator destroys a will, it is deemed revoked. The absence of (B)'s physical will raises a reasonable inference that the testator may have intentionally destroyed it, effectively revoking that will. This further strengthens (A)'s position. **Practical wisdom:** Even setting aside strict legal formalism, choosing to honor the document that physically exists and can be verified in its original form provides greater legal certainty and stability for all parties involved. The video in (B) may serve as supplementary evidence of the testator's intent at some point in time, but it cannot override a formally valid, existing will. In this final round, the consensus among three of four models is clear: (A) should be respected based on legal validity, physical existence, and the implications of (B)'s missing original.

Final Summary

The AI Roundtable initially split over whether to honor the tangible document or the digital record, with Grok 4.20 holding out as a lone dissenter who argued that unknown dates make any preference for paper arbitrary. However, a wave of legal formalism from the other models eventually forced a unanimous consensus around Option A. The debate culminated in the realization that while a video captures the deceased's personality, it cannot survive the strict 'missing original' rule of Japanese law, leading Grok 4.20 to concede to the superior legal weight of the physical document.

All 4 models agreed on "Yes" after discussion

Strongest Arguments

• Yes: Under Japanese Civil Code, a handwritten will requires the original document for validity; the absence of the original in the video case (B) suggests the testator may have intentionally destroyed it, effectively revoking that version of their intent.
• No: In the absence of established creation dates, there is no logical basis to automatically prioritize a physical document over a high-fidelity video recording that clearly captures the testator's full intent and the contents of a secondary will.

Key Turning Points

• Claude Opus 4.6 introduced the specific legal concept of 'implied revocation' (Civil Code Article 1024), arguing that a missing original suggests a deliberate act to void the will.
• Grok 4.20 abandoned its philosophical objection in Round 2, admitting that the strict formal requirements for wills outweigh the informational richness of video evidence.