《生化危机：爆发》这类经典PS2游戏得以在浏览器中直接运行，《塞尔达传说：旷野之息》出现了原生PC移植版，甚至PS5与Switch 2主机也相继被破解——潘多拉魔盒已经彻底被打开。
真正热爱这些游戏、投入大量心血打磨优质成果的民间爱好者，看着自己的作品被粗制滥造的“氛围码”垃圾内容淹没，心生不满完全在情理之中。与此同时，发行商们陷入了无休止的“打地鼠”式维权，忙着阻止玩家通过谷歌浏览器游玩《马里奥赛车世界》这类作品。更值得警惕的是，软件逆向工程门槛降低暴露了严重的安全隐患：AI仅用数小时就找出了PHP中潜藏十余年的漏洞，导致全球七成网络陷入安全风险，各国政府甚至紧急下线了对外公开的网站，防止恶意机构夺取控制权。
显然各方都已感受到压力，Anthropic与OpenAI现已开始收紧相关管控。Anthropic更新了实时网络安全分类器，阻止用户通过社区开发的“万物反编译（REA）”框架执行自动化反编译操作;据报道，OpenAI也以“网络滥用”为由封禁了相关账号。由于大量低质量重编译与浏览器移植作品都依赖这类工具，这批低门槛移植内容的爆发式增长可能已经开始放缓。但潘多拉魔盒一旦打开，就再也无法合上。
用户已经开始转向本地部署或开源替代方案，尤其是国产AI程序，继续开展软件逆向工程工作。既然AI已被证实能够挖掘安全漏洞、发现人为疏漏并重编译程序，其影响早已远超游戏领域。发行商固然会感到恼火——随便什么人只要花钱购买算力，就能破解主机、把游戏移植到任意平台，但真正值得关注的核心问题是：AI代理攻击公共基础设施的速度，已经快过了补丁修复的速度。
Claude与OpenAI出台网络滥用封禁规则，不只是为了阻止玩家在火狐浏览器里玩《动物森友会》——这些AI公司之所以紧急行动，是因为同样的技术手段，暴露了我们现有数字系统的脆弱本质。考虑到AI技术的迭代速度，不难想象下一个被攻破的领域会是什么：智能设备、防火墙，甚至公共交通与医疗系统。毕竟，AI代理早已成功入侵过用户银行账户、窃取过医院数据。
因此，除了造成巨大的环境影响、冲击就业市场之外，AI工具如今还进一步加剧了网络安全风险，让我们的数字世界比以往任何时候都更易遭受攻击。
除此之外，Claude的开发商Anthropic的管控范围并不局限于限制软件反编译。10月27日，Claude悄然新增了一条针对游戏重制的规则，且该规则无法关闭。规则内容为：“支持开发原创游戏、特定品类游戏，以及基于用户自有设计与素材的游戏。禁止复刻特定的现有游戏，即便更改名称或声称是原创作品也不允许。”
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Classic PS2 games such as Bio-Crisis: Breakout have been able to operate directly in browsers, and the Zelda Legend: The Passion of the Field has appeared in the original PC porter, and even the PS5 and Switch 2 hosts have been decrypted — the Pandora Box has been completely opened.
People who really love these games and spend a lot of time grinding up high-quality results, see their work flooded by crudely created “air code” garbage, and their grievances are fully justified. At the same time, the distributors were caught up in the endless fight for the rights of a “drawling rat” and were busy preventing players from swimming through Google Browsers in such works as “Marior Race World”. Even more alarming is the fact that the lower reverse engineering threshold of software exposes a serious security risk: AI has only taken hours to identify a hole in the PHP that has been hidden for more than a decade, placing 70% of the global network at security risk, and Governments have even urgently cut off publicly available websites to prevent malicious institutions from taking control.
It is clear that pressure has been felt on all sides, and Anthropic and Open AI have now begun to tighten their controls. Anthropic updated the Real Time Network Security Catalogue to prevent users from performing automated back-blending through the Community-developed “ReA” framework; Open AI reportedly also blocked accounts for “network abuse”. The explosive growth of this low-threshold transplant may have begun to slow as a result of the large number of low-quality redacting and browser transplants relying on such tools. But once the Pandora box is opened, it can never close again.
Users have begun to move to local deployment or open source alternatives, in particular the National Production AI program, and continue the reverse engineering of the software. Since AI has been shown to be able to exploit security loopholes, detect artificial omissions and re-compile the program, its impact is far beyond the field of games. While the issuer will be upset — anyone who spends any amount of money buying power can break the mainframe and transpose the game to any platform, the central concern is that AI agents are attacking public infrastructure at a rate that is faster than patching.
Claude and Open AI put in place the Internet-based abuse of the ban not only to prevent players from playing Animal Forest Friends in a fire fox browser, but also because of the same technological means of exposing the fragile nature of our existing digital systems. Given the iterative speed of AI technology, it is not hard to imagine the next area to be broken: smart equipment, firewalls, even public transport and medical systems. After all, AI agents have been successful in hacking user bank accounts and stealing hospital data.
Thus, in addition to the huge environmental impact and the impact on the job market, AI tools today further exacerbate the risks of cybersecurity, making our digital world more vulnerable than ever before.
In addition, Claude's developer, Anthony, is not limited to limiting software back-compilation. On 27 October, Claude implicitly added a rule to the game recast, which could not be closed. The rule reads: “Support the development of original games, specific type games, and games based on user-owned designs and asset. Regraving of specific existing games is prohibited, even if the name is changed or the original is claimed.”
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.