激励前置与梯度赋权:生成式人工智能提示词保护的初步设想

Ex-Ante Incentives and Gradient Empowerment:A Preliminary Framework for the Protection of Generative AI Prompts

中文关键词:  提示词  可版权性  因果关系  激励前置  梯度赋权

英文关键词:prompts  copyrightability  causal attribution  ex-ante incentives  gradient empowerment

中文摘要:

      提示词是用户为引导人工智能创作而精心设计的输入文本,是指导人工智能模型进行创作的专属语言,其作为一种可以被识别的“表达”与生成物共存。基于提示词的生成式人工智能创作模式在算法黑箱下模糊了人机贡献界限,传统的“工具论”低估了人工智能的“智能性”,机械地将版权保护重心落在生成物上,对生成式人工智能的版权问题不能精准回应。“贡献论”虽注意到提示词在生成式人工智能运作中的重要性,但是未能进一步区分人机贡献,混淆了提示词的独创性与生成物的独创性的关系,片面将提示词的价值维系于生成物之上,忽略提示词在人工智能内容输出中的应有价值。随着提示词交易市场的形成壮大,激励提示词创作者利用人工智能进行“创作”应成为制度设计的重点。在证成提示词可保护性的基础上,针对结果导向保护模式的固有缺陷,通过“激励前置”与“梯度赋权”方式为提示词提供保护是一种切实可行的规制路径。

英文摘要:

      Prompts are input text carefully crafted by users to guide artificial intelligence (AI) creation, functioning as specialized linguistic instructions for directing AI models. As an identifiable expression, prompts coexist with the generated content they help produce. The generative AI creation model based on prompts blurs the boundary between human and machine contributions under the opacity of algorithmic “black boxes.” The traditional instrumental theory underestimates the “intelligence” of AI and mechanically shifts the focus of copyright protection solely to the generated content, failing to adequately address the copyright issues arising from generative AI. While the contribution theory recognizes the importance of prompts in the operation of generative AI, it does not further differentiate between human and machine contributions. It conflates the originality of prompts with that of generated outputs and narrowly ties the value of prompts to that of generated outputs, overlooking the intrinsic value of prompts in AI-generated content. As the prompt trading markets continue to grow, driving prompt creators to engage in AI-driven “creation” should become a key focus of institutional design. On the premise that prompts are protectable, addressing the inherent shortcomings of result-oriented protection models by adopting approaches such as establishing ex-ante incentive mechanisms and gradient empowerment frameworks is a practical and feasible regulatory pathway for the protection of prompts. 