据大厂日爆爆料，字节跳动近日出台新的内部AI工具使用管理规定，将大范围限制员工通过个人账号使用Claude、Codex等外部AI工具访问公司内部资料。据悉，自本月起，字节跳动将对全公司所有L4级别的代码仓库和飞书文档实施外部AI访问封禁，员工自行购买的AI账号将无法继续访问相关内部资源。
这一通知于上周在字节跳动内部发布，随后引发部分员工讨论。相关讨论主要围绕外部AI工具的使用权限、内部信息安全，以及公司自有AI产品与第三方工具之间的关系展开。
针对员工的疑问，字节跳动相关负责人洪定坤在公司内网作出详细回应。他强调，内部信息安全是此次调整的重要考虑因素，“L3及以上内容一旦有泄露，后果是非常严重的”。
对于此次限制是否意在推动员工使用公司自有AI产品，洪定坤也作出回应。他表示，“推广产品不需要用这种禁掉其他选择强迫使用的方式，因为这样效果很差”。
近年来，Claude、Codex等AI编程工具逐渐成为开发者日常工作的重要组成部分。部分开发者通过个人订阅账号使用相关服务，以辅助代码编写、调试和项目开发。不过，当外部AI工具需要读取企业内部代码仓库、技术文档等资料时，也可能带来敏感信息向第三方服务传输的风险。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
According to daily explosions at large plants, byte has recently introduced new regulations for the use of internal AI tools, which will limit the use of external AI tools such as Claude, Codex and other personal accounts on a wide scale. It is understood that since this month, the byte jump will impose an external AI access ban on all company-wide L4-grade code stores and flight book files and that the AI account numbers purchased by staff will not allow continued access to relevant internal resources.
The announcement was made last week at byte inside the beat publish, which was followed by some staff discussions. Discussions focused on the use of external AI tools, internal information security and the relationship between corporate AI products and third-party tools.
In response to questions from employees, byte beats the responsible person, Hong Dingkun, to provide a detailed response on the inside of the company. He stressed that internal information security was an important consideration in this adjustment and that “the consequences of leaking L3 and above are very serious”.
Hong was also responding to whether the restriction was intended to promote the use of company-owned AI products by employees. He stated that “the promotion of products does not require this ban on other options for forced use, which is not effective”.
In recent years, AI programming tools such as Claude and Codex have become an important part of the day-to-day work of developers. Some developers use related services through individual subscriptions to accounts to prepare, debug and develop projects in support code. However, when external AI tools require access to information such as enterprise in-house code warehouses, technical files, etc., there may also be a risk of sensitive information being transmitted to third-party services.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.