AI实验室从破产初创公司购买数据训练模型

一个新兴的数据交易市场正在形成，AI实验室通过专门的服务商从破产初创公司购买代码和工作区数据用于模型训练。SimpleClosure和Sunset等服务商协助这些初创公司出售电子邮件、聊天记录和文档等数据。1

SimpleClosure在4月添加了数据销售服务，已处理超过150笔交易，其首席执行官Dori Yona表示47%的交易已包括代码或数据销售。1个别创业者通过这种方式获得了可观收入——有人通过销售代码获得约25000美元，通过工作区数据交易获得约15000美元。1SimpleClosure预计到2026年底将向初创公司返还超过1000万美元。1这一现象也吸引了大型科技公司的参与。谷歌在Spirit航空公司破产时出价1000万美元购买其数据，包括1亿封电子邮件、5亿条Teams消息和176000条员工记录。1

然而，这一交易市场面临严重的隐私和伦理问题。法律顾问Greg Ewing称这类数据"在法律上具有放射性"。1法律教授Ifeoma Ajunwa指出员工数据销售前应获得明确同意，但联邦法律目前未禁止此类销售。1更大的问题在于数据收益通常流向投资者而非受其影响的员工。1

A nascent market has emerged in which artificial intelligence laboratories purchase data from bankrupt startups to fuel model development.1 Services such as SimpleClosure and Sunset facilitate these transactions by helping foundering companies sell source code and workplace data—including emails, chat messages, and documents—to interested buyers.1 SimpleClosure added data sales capabilities in April and has already processed more than 150 transactions, with Chief Executive Dori Yona reporting that 47 percent of deals now include code or data sales.1 One entrepreneur obtained approximately $25,000 from selling code, while workspace data transactions have generated around $15,000.1

The scale of these acquisitions reflects significant demand from major technology firms.1 Google bid $10 million to purchase data from Spirit Airlines' bankruptcy proceedings, acquiring a collection encompassing 100 million emails, 500 million Teams messages, and 176,000 employee records.1 SimpleClosure estimates it will return more than $10 million to startups by the end of 2026, while competitors like Sunset are also experiencing rapid growth in data resale operations.1

However, the practice raises substantial privacy and ethical concerns.1 Legal advisor Greg Ewing characterized such data as "radioactive from a legal standpoint," and legal scholar Ifeoma Ajunwa stressed that employees should provide explicit consent before their data is sold.1 Federal law currently does not prohibit these sales, though the revenue typically flows to investors rather than employees themselves.1