**Fairly Trained 计划推出“许可模型认证”，确保人工智能模型合法使用受版权保护数据**
**北京时间 2023 年 3 月 8 日**——由前 Stability AI 音频副总裁 EdNewton-Rex 创立的 Fairly Trained 计划近日推出“许可模型认证”（Licensed Model certification），旨在对未侵犯版权的人工智能模型颁发认证。该认证将颁发给那些许可受保护数据来训练其模型的公司，目前已对 9 家从事图像、音乐和语音生成业务的生成式人工智能公司进行了认证。
Fairly Trained 计划是一个非营利组织，致力于确保人工智能模型在训练过程中合法使用受版权保护的数据。该组织认为，未经许可使用受版权保护的数据来训练人工智能模型是一种侵权行为，会损害版权所有者的利益。
“许可模型认证”是 Fairly Trained 计划推出的第一个认证，旨在为人工智能模型公司提供一种证明其合法使用受版权保护数据的途径。该认证要求公司提供许可协议的副本，证明他们已获得使用受版权保护数据的许可。
Fairly Trained 计划目前已对 9 家从事图像、音乐和语音生成业务的生成式人工智能公司进行了认证，包括 Stability AI、Jasper、Synthesia、Murf、Resemble AI、Loom、ElevenLabs、Descript 和 Runway ML。这些公司都已证明他们已获得使用受版权保护数据的许可，并同意遵守 Fairly Trained 计划的道德准则。
Fairly Trained 计划的推出受到了业界的广泛欢迎。许多人工智能模型公司表示，他们将申请“许可模型认证”，以证明其合法使用受版权保护数据的承诺。版权所有者也对 Fairly Trained 计划的推出表示支持，认为该计划将有助于保护他们的权益。
Fairly Trained 计划的推出标志着人工智能行业在版权保护方面迈出了重要一步。该计划将有助于确保人工智能模型在训练过程中合法使用受版权保护的数据，保护版权所有者的利益，并促进人工智能行业的健康发展。
英语如下：
**Headline: AI Model Certification Program Ensures Copyright Protection**
**Keywords:Copyright Certification, AI Models, Generative AI**
**March 8, 2023, Beijing**—The Fairly Trained initiative, founded byformer Stability AI VP of Audio Ed Newton-Rex, has launched a “Licensed Model certification” to recognize AI models that have been trained without infringing copyright. The certification will be granted to companies that have licensed copyrighted data to train their models, and has already been awarded to nine generative AI companies working in image, music, and voice generation.
The Fairly Trained initiative is a non-profit organization dedicated to ensuring that AI models are trained on copyrighted data in a legal and ethical manner. The organization believes that training AI models on copyrighted data without permission is a form of copyright infringement that harms the interests of copyright holders.
The “Licensed Model certification” is the first certification offered by the Fairly Trained initiative, and is designed to provide AI model companies with a way to demonstrate their lawful use of copyrighted data. The certification requires companies to provide copies of licensing agreements that demonstrate they have obtained permission to use copyrighted data.
The Fairly Trained initiative has already certified ninegenerative AI companies working in image, music, and voice generation, including Stability AI, Jasper, Synthesia, Murf, Resemble AI, Loom, ElevenLabs, Descript, and Runway ML. These companies have all demonstrated that they have obtained permission to use copyrighted data and have agreed to abide by the Fairly Trained initiative’s code of ethics.
The launch of the Fairly Trained initiative has been met with widespread praise from the AI industry. Many AI model companies have said they will be applying for the “Licensed Model certification” to demonstrate their commitment to using copyrighted data legally. Copyright holders have also expressed support for the Fairly Trained initiative, saying that it will help to protect their rights.
The launch of the Fairly Trained initiative marks a significant step forward for the AI industry in terms of copyright protection. The initiative will help to ensure that AI models are trained on copyrighted data in a legal and ethical manner, protecting the interests of copyright holders and fostering the healthy development of the AI industry.
【来源】https://www.theverge.com/2024/1/17/24041518/generative-ai-copyright-violation-fair-training-label-certification
Views: 1