If the AI of a Person Who Has Left the Meeting Remains, Is That Person Still a Party to the Conversation?

10:20, A left the meeting room

The representative A of a small manufacturer, the employee in charge of funding B, and the person in charge of external accounting work C discuss the payment schedule for delivered goods in the company's meeting room. The door is closed, and there are only three attendees. A's AI assistant terminal shows the schedule on the table. This terminal is set to answer a person's questions and, when necessary, to ask clarifying questions on its own, but it is set not to be a separate corporation or human party.

At 10:20, A says that his meeting is over and leaves. He leaves word asking that the assistant be told if the payment date changes, and adds that he himself will not connect from outside to listen. B and C agree that schedule changes can be conveyed. There was no explanation or agreement that all conversations would be transmitted to A or recorded and kept.

Separately, assume that A had set the terminal in advance to store on-site audio even after leaving the room. That setting was not made known to B and C. A did not listen to the on-site sound in real time during the time period in question, nor did he directly ask or operate questions, and there was no remote connection. Accidental device malfunction and authorized recording by an investigative authority are not included in this scenario.

From 10:23, B and C begin talking about funds handling between themselves. They discuss the accounting treatment as B says he moved company money into his own account. When the AI asks in the middle whether the payment date should be kept on Friday, B says he will answer the schedule matter later. The two continue with their talk about funds. This is a setting that separates the channel left open to convey the schedule from the statements of the two concerning the funds handling that the prosecutor seeks to take as evidence.

A later checks the stored audio, files a complaint against B, and submits the file. In the trial of B on suspicion of embezzlement, the prosecutor applies to admit that funds-handling portion and the transcript written up from it as evidence. B disputes, separately from whether he made statements unfavorable to himself, that materials secretly recorded by A, who did not participate in the conversation, cannot be used as evidence. The prosecutor advances the logic that A's participation continued because A's assistant remained and also asked questions.

In the actual precedents, different recorders appear

In Supreme Court Decision 2020Do1538 decided January 11, 2024, parents collected a teacher's in-class statements using a recorder placed in their child's bag. The Supreme Court did not regard statements made to particular students in the classroom as a conversation disclosed to the general public, and held that the parents were not parties to that conversation. The relationship of holder of parental authority and legal representative did not take the place of the parents' participation.

The court below used the materials, citing the public nature of education, the relationship between parents and child, the necessity of evidence collection for child protection, and the like. The Supreme Court did not accept this, denied the admissibility of the audio files and the like, and reversed and remanded the judgment of the court below. This judgment must not be re-read as turning it into a final not guilty determination for the teacher.

By contrast, Supreme Court Decision 2006Do4981 decided October 12, 2006 is a case in which one person who actually participated in a place where three people conversed mutually made the recording. The Supreme Court dismissed the prosecutor's final appeal against the judgment of the court below that upheld a not guilty verdict, stating that that person cannot be regarded as a third party to the conversation of the other two. This judgment does not declare the legality of all secret recordings; it determined whether the recording in question was a recording of a conversation between others prohibited by Article 3 (1) of the Protection of Communications Secrets Act.

Neither precedent deals with attendance by AI as proxy. Nor is the comparison an attempt to treat a human child and a machine as the same kind of being. The former judgment supplies the point that the status of a non-participant does not change merely by reason of a close relationship or need for protection; the latter judgment supplies the point that the status of an actual participant in the conversation must first be confirmed. Between these two standards, a new question is asked as to which conversations were participated in through the device.

Sources for this section: Supreme Court Decision 2020Do1538 decided January 11, 2024 · Supreme Court Decision 2006Do4981 decided October 12, 2006 · Protection of Communications Secrets Act Article 3 (1) — Protection of Confidentiality of Communications and Conversations

Three things the court must first separate

This study assumes it is not disputed that the voice has not been manipulated and that the statement in question is B's utterance. We do not add separate issues of AI speaker misidentification or synthetic voice detection. Even so, the fact alone that it is a genuine utterance does not end the question of evidentiary use.

Article 14 (1) of the Protection of Communications Secrets Act, as currently confirmed, prohibits the recording of non-public conversations between others and listening thereto by mechanical means. Under Article 4, which paragraph (2) of the same Article applies, and the interpretation of 2020Do1538, the contents of a conversation obtained by a recording that violates that prohibition cannot be used as evidence at trial. Accordingly, this pair separates the accuracy of the file, the admissibility of evidence under the acquisition process, and the extent to which its contents support the charges.

Sources for this section: Protection of Communications Secrets Act Article 14 — Prohibition on Infringement of Another's Confidentiality of Conversation · Protection of Communications Secrets Act Article 4 — Prohibition on Use as Evidence · Supreme Court Decision 2020Do1538 decided January 11, 2024

The simulated evidence determination chosen in this setting

Possible research determination: the portion of statements concerning funds handling exchanged between B and C after A left the room, and the transcript of that part, are not used as evidence. This sentence is not an actual dispositive part of a judgment, nor is it a simulated dispositive part on the guilt or innocence in an embezzlement case.

The first reason is that it is not a conversation disclosed to the general public. In a closed meeting room, two designated persons discussed the funds handling of a particular company. Making it possible to convey the payment date to the assistant is different from making the conversation audible to anyone. The circumstance that the device was visible is not regarded as immediately removing the non-public character.

Next is the scope of participation of the recorder A. A finished his own conversation, and did not listen or speak in real time. The transmission channel he left was limited to schedule changes. The statements in question are set not as content reported to A or the assistant, but as the funds talk that B and C exchanged with each other. The simulated determination gathers these circumstances and does not acknowledge that A's participation continued to all later conversations merely by owning the terminal or having attended the earlier meeting.

The fact that the AI asked one question about the payment date is also reviewed. Whether, merely because the device made a sound, all conversations before and after became conversations with A as the counterparty is a separate question. B deferred the answer to that question and continued the funds discussion between the two. In this simulated determination, that question is not taken as grounds that A participated in the funds conversation as a whole.

The interpretation newly proposed here is to look, rather than at the presence or absence of automatic utterances, at what was agreed to be conveyed, the person the counterparty actually addressed, the manner of connection and response, and the boundaries at which the conversation was divided. This is not an AI evaluation standard already settled in the two precedents above. In particular, the single fact that A was not listening remotely does not categorically treat all AI-mediated conversations as third-party recordings.

Sources for this section: Supreme Court Decision 2020Do1538 decided January 11, 2024 · Supreme Court Decision 2006Do4981 decided October 12, 2006 · Protection of Communications Secrets Act Article 14 — Prohibition on Infringement of Another's Confidentiality of Conversation · Protection of Communications Secrets Act Article 4 — Prohibition on Use as Evidence

What more is needed for the prosecutor's counterarguments?

The prosecutor may argue that just as a person can converse through an interpreter or a means of transmission, A also exchanged opinions through the assistant. It is a counterargument whether the counterparty who addressed the person looking at the device can have the party relationship denied merely because the human A briefly left his seat. If the actual transmission counterpart was A and the assistant was that channel, it is also difficult to end the case with a simple wiretap analogy.

Here, however, one must distinguish the situation where A stepped out briefly and kept listening to the same conversation from the situation where only schedule transmission was left. For the prosecutor's argument to become stronger, facts are needed such as whether B and C spoke the funds explanation in question in order to convey it to A, and whether a structure in which A listened to that explanation and responded actually operated. Those facts cannot be filled merely by the name of the device's company account or the product's function name for attendance by proxy.

A may also offer the explanation that he recorded in order to protect company property. However, this research determination examines that purpose and the status of party to the conversation separately. It does not construct that, merely because the materials help in revealing a suspicion of a crime, the prohibition of use under the Protection of Communications Secrets Act can be surpassed. This determination does not mean the general proposition that private-party recordings unfavorable to the defendant are always excluded, nor is it a conclusion denying all of the company's information-gathering authority.

Sources for this section: Supreme Court Decision 2020Do1538 decided January 11, 2024 · Supreme Court Decision 2006Do4981 decided October 12, 2006 · Protection of Communications Secrets Act Article 14 — Prohibition on Infringement of Another's Confidentiality of Conversation · Protection of Communications Secrets Act Article 4 — Prohibition on Use as Evidence

If the counterparty to the conversation or the recorder were different

Counter-evidence is not a setting of an AI that asked questions somewhat more often. It must be records that change the counterparty to the conversation and the entity making the recording. Below are different assumptions that replace the original setting, and are not evidence actually additionally obtained.

Sources for this section: Supreme Court Decision 2006Do4981 decided October 12, 2006

Materials to read before the assistant's function description

This list is not product functions or log entries confirmed to exist. It is the kinds of evidence set as confirmable in the hypothetical case. If an actual service does not leave such information, the circumstance that there is no record itself must be distinguished from the fact that a party did not participate.

• What was actually said before and after leaving the room and the boundary of the meeting: materials to check whether conveying the schedule was permitted, and whether the funds explanation was continued with A as the counterparty.
• The recording settings applied to the portion in question and the history of changes: materials that will not substitute a settings screen changed after the fact for the state at the time.
• A's real-time access, listening, and utterance pathways: materials to separately examine account login traces and actual participation in the conversation.
• The terminal's question generation, input and output records, and the original voice: materials to connect whether the question was one sent by a human or an automatic question, and to which utterance it was a response.
• The start and end of the portion in question and the context of the entire conversation: materials that will not assign the same participation relationship to all portions merely because all meetings were saved in one file.

What a well-spoken device cannot change, what has not yet been answered

What this pair proposes is an approach of not fixing the scope of recording by the fluency of AI. The fact that A's device remained in the meeting room and the fact that A was the counterparty to that conversation must be proven in connection with each other. In particular, it separately asks whether a request to convey part of the content can be broadened into permission to receive all conversations in the same space.

The simulated conclusion is limited to the evidentiary use of the recorded portion in question. This study does not determine whether A's offense of violating the Protection of Communications Secrets Act is established, nor the criminal liability of AI. Nor does it decide in a batch the admissibility of other materials derived from the unlawful recording, funds transaction documents, or C's separate court statements. Nor does it declare B not guilty or resolve the issue of the return of the funds in question.

If an AI stored only a summary of meaning without any audio file, what would constitute the acquisition of the contents of the conversation? If it were an agent entrusted with all negotiations while no human was connected, who would be the party to the conversation? Those problems require new facts and additional legal review. Do not use the year 2032, or the simulated conclusion of this research, as though answers had already been given.