OpenAI、ChatGPT上の中国の影響工作を暴露 — 一方Anthropicは競合モデル研究を復活
OpenAIはchatgptを利用した中国の影響工作キャンペーンを検出。一方、anthropicは競合モデルに関する研究を制限するポリシーを撤回した。これらの動きは、激化する地政学的な攻防と、オープンなAI研究をめぐる内部論争を示唆している。
openai revealed it disrupted a chinese influence operation utilizing chatgpt to produce content focused on us politics, russia, and the ukraine conflict, part of a wider sweep targeting propaganda operations. concurrently, anthropic updated its acceptable use policy to explicitly allow security researchers to test competing models, walking back restrictions that were widely condemned as anti-competitive. anthropic's initial policy had threatened to revoke access from researchers generating data for rivals, a practice standard in open-source evaluation. the reversal follows significant backlash from the ai research community regarding transparency and scientific freedom. these events underscore the dual pressure on ai labs to mitigate state-backed disinformation while maintaining trust with the research ecosystem.