创建

登录 ReadmeX

或

了解我们如何处理你的信息: 隐私政策

ReadmeX

陪你读懂 AI 动态

今天,想先了解什么?

选一份动态总结,或直接聊聊你的问题。

社区和关注动态总结只属于你自己。

资讯

OpenAI封号,Claude限制逆向协助

AI 总结

新智元报道,安全研究者@log4j在本地逆向分析自家三星电视后,其ChatGPT账号被OpenAI以“网络滥用”为由封禁。@log4j称自己未利用OpenAI挖掘漏洞或攻击他人,其所称约500美元损失也未获独立证实;报道还援引开发者称,Claude曾中止或拒绝与开源REA工具相关的逆向协助。

深度见解

据报道,即使研究者声称工作仅涉及自有设备且未攻击第三方,AI服务的安全政策仍可能直接影响逆向工程工作流。

Reverse-engineering games for converting them between systems is currently a hot topic, and it's getting messier by the day. According to multiple user claims, Anthropic's Claude Code and/or Opus 5.5 model are blocking attempts at innocuous reverse-engineering, including but not limited to those using the popular Reverse Engineer Anything (REA) tool. That issue may be linked to a text rule that's now in force, which allegedly instructs the model to avoid "[recreating] a specific existing game." According to Twitter user Chasm, the game-related textual rule is "Do build original games, games in a genre, and games from the user’s own designs and assets. Don’t build a recreation of a specific existing game, even if it is renamed or called original." There is no official or additional confirmation if this is the actual rule text, but it's plausible enough given the very clear refusals that devs have been encountering over the past week.

This situation is tricky to analyze due to the varying types of reports. First, the existence of a game-related rule is real, as Claude Code's GitHub is seeing an influx of issues complaining about refusals to do standard game work halfway through development, when replicating a game mechanic, and patching a tabletop-like game. It's possible that some of those reports have been made by users with nefarious purposes, but at least one is a bona fide developer with a real track record.

#OpenAI#ChatGPT#Anthropic#Claude#Reverse Engineer Anything (REA)#Claude Code#Opus 5.5#Samsung

原文约 3 分钟读完阅读原文
群聊

还没有评论,来说说你的看法。