Skip to main content
Podcast Episode: Securing Journalism on the ‘Data-Greedy’ Internet

New technologies are radically advancing our freedoms, but they are also enabling unparalleled invasions of privacy. National and international laws have yet to catch up with the evolving need for privacy that comes with new digital technologies. Respect for individuals' autonomy, anonymous speech, and the right to free association must be balanced against legitimate concerns like law enforcement. EFF fights in the courts and Congress to maintain your privacy rights in the digital world, and works with partners around the globe to support the development of privacy-protecting technologies.

Your cell phone helps you keep in touch with friends and family, but it also makes it easier for the government to track your location.

Your Web searches about sensitive medical information might seem a secret between you and your search engine, but companies like Google are creating a treasure trove of personal information by logging your online activities, and making it potentially available to any party wielding enough cash or a subpoena.

And the next time you try to board a plane, watch out—you might be turned away after being mistakenly placed on a government watch list, or be forced to open your email in the security line.

Several governments have also chosen to use malware to engage in extra-legal spying or system sabotage for dissidents or non-citizens, all in the name of “national security.”

As privacy needs evolve, so too should our regulatory regimes. National governments must put legal checks in place to prevent abuse of state powers, and international bodies need to consider how a changing technological environment shapes security agencies’ best practices. Above all, we need to respect the rights of autonomy, anonymity, association, and expression that privacy makes possible, while also taking into account legitimate law enforcement concerns.

Read our work on privacy issues below, and join EFF to help support our efforts.

For information about the law and technology of government surveillance in the United States check out EFF's Surveillance Self-Defense project.

Privacy Highlights

Banner Graphic: 

Digital Privacy at the U.S. Border: Protecting the Data On Your Devices and In the Cloud

Digital Privacy at the U.S. Border:Protecting the Data On Your Devices and In the Cloudby Sophia Cope, Amul Kalia, Seth Schoen, and Adam SchwartzDownload the report as a PDF.EXECUTIVE SUMMARYأفادت الحكومة الأمريكية أن عدد حالات تفحص المحتويات الالكترونية على الحدود قد إزداد بمقدار خمسة أضعاف خلال سنة واحدة فقط،...

NSA Spying

The US government, with assistance from major telecommunications carriers including AT&T, has engaged in massive, illegal dragnet surveillance of the domestic communications and communications records of millions of ordinary Americans since at least 2001. Since this was first reported on by the press and discovered by the public in late...

Privacy Updates

A robot painting a self-portrait

我们如何看待版权与AI艺术

可以理解,艺术家们忧心像 Stable Diffusion 这样的自动图像生成器可能会削弱他们艺术作品的市场。 毕竟,我们的社会并不对自动失业提供保障,而作为一名视觉艺术家本已是一个不稳定的职业。 在这种情况下,人们很自然地想到了版权法,因为版权法本应有助于确保艺术家能从他们的作品中获得创作的报酬。 不幸的是,在最近一场由一群艺术家发起的针对 Stable Diffusion 的集体诉讼中,法庭推进了一种对人类创作者来说是极其危险的版权理论。 其他理论——包括在那起诉讼和 Getty Images 的另一起诉讼中的理论——一般建议以干涉研究、搜索引擎以及使新旧技术互动能力的方式改变和扩大版权限制。 本文的法律分析是我们另一篇描述 AI 图像生成技术以及我们如何看待其潜在风险和收益的帖子的姊妹篇。 我们建议您先阅读那篇文章了解上下文,然后再回到这篇文章来了解我们对美国法下版权问题的看法。 总的来说,版权法应在给予艺术家充分的创作激励(通过赋予他们控制其艺术作品的某些使用方式)与赋予公众基于已有作品进行创新或将已有作品使用于新的、有趣的方式的权利之间取得平衡。 这里的问题是,是否图像版权所有人应当有权阻止他人使用其版权图像进行AI生成模拟器训练。 要想回答这个问题,首先我们需要了解一些美国版权法的基本原则。 首先,版权法不禁止对作品中所体现的事实进行观察或复制(这称为“思想/表达区别”)。 相反,版权禁止以可以替代原作的方式对原作的创造性表达进行复制,以及过度使用原作的创造性表达进行“衍生作品”制作。 其次,即使有人复制或创作衍生作品,如果是“合理使用”,则该使用不构成侵权。 使用是否公平取决于多种因素,包括使用目的、原创作品的性质、使用量以及对原创作品市场的潜在危害。 版权与训练集 步骤1: 从网络抓取图像 与复制以创建搜索引擎或其他分析用途一样,下载图像以对其进行分析和索引以创建新的非侵权图像很可能属于合理使用。 当一项行为可能涉及版权但又是实现非侵权使用的必要步骤时,它通常本身就属于合理使用。 毕竟,以非侵权方式使用作品的权利只有在非版权人也被允许执行这些创作的必要步骤时才有意义。 因此,作为中间用途和分析用途,抓取不太可能违反版权法。 步骤2: 存储图像相关信息 在此步骤中,系统分析图像并存储有关像素排列如何与文本注释中的单词相关联的信息。 Stable Diffusion模型对超过 50 亿张图像进行了 4 GB 的观察。 这意味着该模型所含有的对于每个被分析图像的信息少于一个字节(一个字节只有八位——一个零或一个一)。 针对 Stable Diffusion 的诉状将此过程描述为“压缩”(并存储)训练图像,但这是错误的。 除了少数例外,基本无法根据存储的有关图像的事实信息重新创建模型中使用的图像。 即使是最小的图像文件也包含数千字节;...

students use books and tablets to hide from a spying eye

After Students Challenged Proctoring Software, French Court Slaps TestWe App With a Suspension

In a preliminary victory in the continuing fight against privacy-invasive software that “watches” students taking tests remotely, a French administrative court outside Paris suspended a university’s use of the e-proctoring platform TestWe, which monitors students through facial recognition and algorithmic analysis.TestWe software, much like Proctorio, Examsoft, and other...

Pages

Back to top

JavaScript license information