为保护艺术家免受人工智慧训练影响而建立的图像分享平台 Cara，近期遭遇了三次大规模的数据抓取事件。尽管平台努力阻挡人工智慧，抓取者仍利用其漏洞窃取了数百万张图片、连结和元数据，这不仅引起了平台一百五十万名用户的恐慌，也迫使创办人张晶娜不得不筹集资金以应对潜在的法律诉讼。（关键数字：1.5）

令人意外的是，第一位抓取数据的学生「Heft」在意识到自己的行为对艺术家社群造成深刻的情感伤害后表达了歉意。他最初只将此视为一项技术测试和在 Reddit 上激怒他人的恶作剧，但后来他主动道歉并加入了 Cara 的团队，协助找出并解释平台在防止抓取方面的结构性弱点。

在体认到没有任何网站能完全防范数据抓取后，张晶娜与 Heft 目前正合作开发一款名为「Lantern」的开源工具。这项工具能让艺术家为其作品建立「单向指纹」，一旦他们的作品出现在公开的人工智慧数据库中，系统就会发送通知，让艺术家在缺乏完善法规的情况下，有机会要求移除作品并重新获得部分控制权。

Cara, an image-sharing platform built to protect artists from artificial intelligence training, recently suffered three major data scraping incidents. Despite the platform's efforts to block AI, scrapers exploited its vulnerabilities to steal millions of images, links, and metadata, causing panic among its 1.5 million users and forcing founder Jingna Zhang to raise funds for potential legal battles.

Surprisingly, the first scraper, a student named "Heft," expressed regret after realizing the deep emotional harm his actions inflicted on the artist community. Originally treating the scrape as a mere technical test and a trolling stunt on Reddit, he later apologized and joined the Cara team to help identify and explain the platform's structural weaknesses in preventing such attacks.

Recognizing that no website is completely immune to scraping, Zhang and Heft are now collaborating on an open-source tool called "Lantern." This tool allows artists to create a "one-way fingerprint" for their artwork so that they can receive notifications if their images appear in public AI datasets, giving them a chance to request removal and regain some control in the absence of comprehensive regulations.