本文来源:https://www.zhiding.cn/ai-applications/2026/1008/3201230.shtml
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
OpenAI推出Dots智能体,具备专属云端计算机
OpenAI在DevDay开发者大会上推出常驻AI智能体Dots,由GPT-6 Astra驱动,每个智能体拥有独立云端计算机,可主动执行任务、跨平台协作,并支持与Slack、Teams集成。目前向ChatGPT Pro及Business Premium订阅用户开放。
OpenAI Group PBC今日在其DevDay开发者大会上宣布推出Dots,这是ChatGPT中一组全天候运行的人工智能智能体。
该发布旨在让ChatGPT用户摆脱一次只能处理一个提示词的使用方式。用户可以给智能体设定一个目标,让它在没有每一步指导的情况下持续推进工作,即便同时还有其他多个项目在进行。每个智能体都会根据反馈学习其所有者的偏好。OpenAI表示,部分工作成果会在用户尚未提出要求之前就已完成。
这些智能体由GPT-6 Astra驱动,每个智能体都拥有自己专属的云端计算机,用户可以随时打开查看其工作状态。通过OpenAI的插件,用户可以连接超过4000个应用程序。在获得授权的情况下,智能体还可以直接在所有者的笔记本电脑上工作。
用户可以在桌面端、网页端或移动端的ChatGPT中与自己的智能体进行消息交流或通话,也可以在Slack和Teams中联系它。由于上下文信息会随智能体在不同渠道间流转,在ChatGPT中启动的项目可以直接转交给Slack中的团队处理,而无需重新说明情况。智能体会将进展更新和问题反馈给所有者。
OpenAI已经在公司内部团队中使用这些智能体。当公司Slack中出现漏洞报告时,智能体会立即展开调查,新的设计方案也会被转化为可运行的应用程序,从而让员工能够专注于处理客户反馈。
一位公司外部的早期测试者曾忘记向某家出版机构开具账单。当他的智能体发现这一情况后,便准备好了发票,并等待他批准后才发送出去。
当没有人给智能体分配任务时,它会主动寻找可以提供帮助的方式,OpenAI将这种行为称为主动研究。在此期间,智能体与所有者应用程序的连接处于只读状态,因此它无法发送消息或更改任何内容。
任何可能涉及用户账户或共享信息的操作,都必须先通过一项名为自动审核的检查,且监控系统可以在出现安全隐患时暂停或终止智能体的运行。OpenAI提醒用户要对重要工作进行审核,因为这些智能体“仍然可能出错”。
OpenAI还预览了专属智能体功能,雇主可以为单一明确的工作任务,为其配置专属身份和凭证。采购和发票处理是OpenAI内部测试的工作任务之一。客户可以从针对性试点项目开始,OpenAI的工程师会协助定义每个智能体的具体职责。微软公司正在与OpenAI合作,以便专属智能体能够通过Agent 365中的治理和安全控制进行管理。
这项新服务也使OpenAI与Meta Platforms Inc.形成竞争,后者于9月8日在美国发布了其个人智能体Muse。与OpenAI的智能体类似,Muse也运行在云端的专属计算机上。Meta表示,大部分使用场景是免费的,公司会向重度用户出售订阅服务。
Dots今日开始向ChatGPT Pro和Business Premium订阅用户推出,部分账户可能需要等待数天才能获得访问权限。欧洲经济区、瑞士和英国的Pro订阅用户不在本次推出范围内。企业客户,包括使用Edu和Healthcare计划的客户,可以在工作区管理员开启后试用测试版。
Pro或Business Premium计划均附带一个免费智能体,无需额外付费。OpenAI的帮助文档显示,智能体的使用量在首月内不会计入计划配额,此后则按各计划条款执行。OpenAI计划在未来提供更多智能体,以及为现有智能体提供更快速度或更高月度使用容量,具体价格尚未公布。
Q&A
Q1:Dots是什么?它能做什么?
A:Dots是OpenAI推出的ChatGPT全天候AI智能体,用户可以给它设定目标,让它自主持续推进工作,无需每一步都进行指导,还能从反馈中学习用户偏好。
Q2:Dots智能体如何保障安全性?
A:涉及用户账户或信息共享的操作必须先通过自动审核检查,监控系统还可以在出现安全隐患时暂停或终止智能体运行,用户仍需审核重要工作结果。
Q3:Dots现在可以在哪里使用?
A:Dots已向ChatGPT Pro和Business Premium订阅用户推出,部分账户需等待数天,欧洲经济区、瑞士和英国的Pro用户暂未包含,企业客户可由管理员开启测试版使用。
Source: https://www.zhiding.cn/ai-applications/2026/1008/3201230.shtml
For copyright issues, please email: 13951431913#139.com
OpenAI Launches Dots Agent with a Dedicated Cloud Computer
At its DevDay developer conference, OpenAI launched Dots, a persistent AI agent powered by GPT-6 Astra. Each agent owns an independent cloud computer, can proactively execute tasks and collaborate across platforms, and supports Slack and Teams integration. Currently open to ChatGPT Pro and Business Premium subscribers.
OpenAI Group PBC announced today at its DevDay developer conference the launch of Dots — a set of AI agents in ChatGPT that run around the clock.
The release aims to free ChatGPT users from the one-prompt-at-a-time way of working. Users can give an agent a goal and let it push the work forward without step-by-step guidance, even while several other projects run in parallel. Each agent learns its owner's preferences from feedback. OpenAI says some work products are completed before the user has even asked.
The agents are powered by GPT-6 Astra, and each one has its own dedicated cloud computer that users can open at any time to check its working status. Through OpenAI's connectors, users can link more than 4,000 applications. When authorized, an agent can also work directly on the owner's laptop.
Users can message or call their agents in ChatGPT on desktop, web or mobile, and can also reach them in Slack and Teams. Because context flows with the agent across channels, a project started in ChatGPT can be handed directly to a team in Slack without re-explaining the situation. Agents send progress updates and questions back to the owner.
OpenAI already uses these agents inside its own teams. When a bug report appears in the company's Slack, the agent immediately starts investigating, and new design proposals are turned into runnable applications — letting employees focus on customer feedback.
An external early tester once forgot to bill a publisher. When his agent noticed, it prepared the invoice and waited for his approval before sending it.
When no one assigns a task, the agent proactively looks for ways to help — a behavior OpenAI calls proactive research. During this time, the agent's connection to the owner's applications is read-only, so it cannot send messages or change anything.
Any operation that might involve the user's account or shared information must first pass a check called automatic review, and a monitoring system can pause or terminate an agent if a security risk appears. OpenAI reminds users to review important work, because these agents "can still make mistakes."
OpenAI also previewed dedicated-agent functionality: employers can configure a dedicated identity and credentials for a single well-defined work task. Procurement and invoice handling are among the work tasks OpenAI is testing internally. Customers can start with targeted pilots, with OpenAI engineers helping define each agent's specific responsibilities. Microsoft is working with OpenAI so dedicated agents can be managed through governance and security controls in Agent 365.
The new service also puts OpenAI in competition with Meta Platforms Inc., which launched its personal agent Muse in the US on September 8. Like OpenAI's agents, Muse runs on a dedicated cloud computer. Meta says most use cases are free and the company will sell subscriptions to heavy users.
Dots starts rolling out today to ChatGPT Pro and Business Premium subscribers, with some accounts possibly needing to wait several days for access. Pro subscribers in the European Economic Area, Switzerland and the UK are not included in this rollout. Enterprise customers, including those on the Edu and Healthcare plans, can try the beta once a workspace admin enables it.
Both the Pro and Business Premium plans include one agent for free, with no extra charge. OpenAI's help documentation shows agent usage does not count toward plan quotas during the first month, after which each plan's terms apply. OpenAI plans to offer more agents in the future, as well as faster speeds or higher monthly usage capacity for existing agents; pricing has not been announced.
Q&A
Q1: What is Dots? What can it do?
A: Dots is OpenAI's around-the-clock AI agent for ChatGPT. Users can set a goal for it to autonomously push work forward without step-by-step guidance, and it learns the user's preferences from feedback.
Q2: How does the Dots agent stay safe?
A: Operations involving the user's account or shared information must first pass an automatic review check; the monitoring system can also pause or terminate the agent if a security risk appears. Users still need to review important work results.
Q3: Where can Dots be used now?
A: Dots is available to ChatGPT Pro and Business Premium subscribers; some accounts need to wait several days. Pro users in the European Economic Area, Switzerland and the UK are not yet included. Enterprise customers can have an admin enable the beta.
