IT之家 10 月 21 日消息,OpenAI 于当地时间周二宣布,使用者现在可以为 ChatGPT 对话上锁。当其他人有可能看到显示屏或使用自己设备时,这项能力可以保护敏感对话的隐私。使用者可以设置四位数字 PIN 码保护锁定对话,在兼容的设备上也可以使用面容 ID(Face ID)或触控 ID(Touch ID)。
据IT之家了解,该能力和应用内的临时对话不同,临时对话属于独立会话,不会浮现在聊天记录里,而锁定对话兼容使用者随时回看。
OpenAI 表示,这项能力面向这类情形:使用者正在讨论私人事项或是策划惊喜,需要把相关对话保护起来,之后把设备交给亲友,向对方展示其他对话里的内容。
随着越来越多人用 ChatGPT 开展私人交流,锁定对话能力为其新增了一层隐私防护。当然,锁定对话虽然可以限制他人访问,但无法应对使用者对于敏感信息存储和处理方式的顾虑。
OpenAI 说明,锁定对话不会启用端到端加密,也不会生成独立加密存储库;不会改变对话数据的留存时长,不会清除锁定前 ChatGPT 已经记住的信息,也不会改动“协助改进模型”的设置。
使用者锁定对话之后,该对话会移入专门的“锁定对话”分区,和普通聊天记录、置顶对话以及聊天搜索结果相互隔离。
如果启动记忆能力,锁定对话分区具备独立的记忆空间,仅在分区内的对话之间共享。锁定之后发送的新消息可以更新这部分独立记忆,但不会改动通用记忆。也就是说,锁定对话产生的新信息不会泄露到普通对话中。
OpenAI 提示,重置 PIN 码会永久删除全部由旧 PIN 保护的锁定对话、文件以及其他内容。
另外,临时对话、归档对话、群组对话、借助分享链接创建的对话、带有定时任务的对话,以及包含会议录音的对话,都无法使用锁定对话能力。
免费版、Go 版、Plus 版、Pro 版、商业版、企业版和教育版 ChatGPT 使用者均可使用锁定对话能力。该能力不适用于医疗专用版 ChatGPT。能力从当地时间周二开始分批上线,并非所有使用者都能立刻使用。
本文来源:https://www.ithome.com/1/015/629.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
OpenAI announced Tuesday, October 21, local time, that users can now lock down ChatGPT conversations. This ability protects the privacy of sensitive conversations when it is possible for others to see the display or use their own devices. Users can set a four-digit pin to protect the locked conversation, or they can use Face ID or Touch ID on a compatible device.
According to IT House, this ability is different from temporary conversations in the app. Temporary conversations are independent conversations and do not appear in the chat log, while locked conversations are compatible with users to watch back at any time.
OpenAI says this capability is for situations where a user is discussing a private matter or planning a surprise and needs to protect the conversation before handing over the device to a friend or relative to show the other person what is going on in the conversation.
As more and more people use ChatGPT for private communication, the ability to lock down conversations adds a new layer of privacy. Of course, locking a conversation can limit access, but it doesn't address the user's concerns about how sensitive information is stored and handled.
OpenAI explains that locking a conversation does not enable end-to-end encryption, nor does it generate a standalone encrypted repository; it does not change the retention period of the conversation data, it does not erase information already remembered by ChatGPT before locking it, and it does not change the setting of Help Improve the Model.
When a user locks a conversation, it moves into a dedicated Lock Conversation section, which is isolated from normal chat history, pinned conversations, and chat search results.
If memory is enabled, the locked conversation partition has a separate memory space that is shared only between conversations within the partition. New messages sent after the lock can update this part of the independent memory, but will not change the general memory. In other words, the new information generated by the locked conversation will not be leaked into the normal conversation.
OpenAI prompts that resetting the pin will permanently delete all locked conversations, files, and other content protected by the old pin.
In addition, the ability to lock down conversations is not available for ad-hoc conversations, archived conversations, group conversations, conversations created through sharing links, conversations with scheduled tasks, and conversations that include audio recordings from meetings.
Free, Go, Plus, Pro, Business, Enterprise, and Education ChatGPT users can use the ability to lock conversations. This capability does not apply to the medical-only version of ChatGPT. Capabilities are rolled out in batches starting Tuesday local time and may not be immediately available to all users.
This article is from: https://www.ithome.com/1/015/629.htm For copyright issues, please write to: 13951431913#139.com or submit a review on the right.