I think itās pretty bad for the AI to write allegations of criminal or objectionable behavior (āTorres harassed and stalked Peter Boghossian...ā etc.) in its own voice, as that exposes the AIās developers to liability for libel if the claims are false. Claims like this should be hedged using phrases like āallegedlyā or āthe author claims that...ā.
Not the botās lawyer, but I might be more worried about universes in which the bot normally does that but doesnāt in a particular case, and the summarized post in that case ends up libelous. In that case, one could argue that the non-use of the typical disclaimers communicated something to the reader.
In contrast, the language here sounds āin its own voice,ā but itās also pretty clear from context that the bot is merely summarizing the original post.
I think itās pretty bad for the AI to write allegations of criminal or objectionable behavior (āTorres harassed and stalked Peter Boghossian...ā etc.) in its own voice, as that exposes the AIās developers to liability for libel if the claims are false. Claims like this should be hedged using phrases like āallegedlyā or āthe author claims that...ā.
Not the botās lawyer, but I might be more worried about universes in which the bot normally does that but doesnāt in a particular case, and the summarized post in that case ends up libelous. In that case, one could argue that the non-use of the typical disclaimers communicated something to the reader.
In contrast, the language here sounds āin its own voice,ā but itās also pretty clear from context that the bot is merely summarizing the original post.