OpenAI、ログ検索で「吐き出し」を確認できる証拠を隠したとして告発される
The New York Times紙は、OpenAIが技術的負担を訴えながら、ログやトレーニングデータから著作権作品を検索する能力を隠していたと主張。OpenAIが法廷で否定する一方で、吐き出し(regurgitation)を追跡するツールを保有していたと伝えています。
the new york times and daily news accused openai of misleading the court regarding its ability to identify copyrighted works in its training data and chat logs. deposition testimony from engineer vinnie monaco contradicted openai's claims, revealing it had built internal search tools and maintained a database of 78 million de-identified conversations. the plaintiffs allege openai violated preservation orders by deleting data and submitting a redacted sample of 20 million logs rendered unusable. they are asking the judge to sanction openai and prevent the company from using the current log samples as evidence. openai denies the allegations, claiming the times is attempting to invade user privacy as its case weakens.