2026年9月29日 / 美国东部时间下午3:09 / 哥伦比亚广播公司新闻
OpenAI周二在其年度开发者大会上发布了一款个人AI助手，加入了越来越多承诺用虚拟助手为消费者生活减负的公司行列。
这家公司的全新AI助手名为“dots”，将协助完成职场及其他各类任务，比如安排会议、预订旅行行程、制定预算、调试电脑系统以及编写软件代码。该助手基于本月早些时候OpenAI发布的新模型GPT‑6 Astra运行。
“这就像一个永远会支持你的AI助手，灵感来源于我们成长过程中在电影里看到的那些酷炫的助手形象，”OpenAI首席执行官山姆·奥特曼在旧金山举行的OpenAI年度开发者大会DevDay的主题演讲中说道。
“你可以将复杂的工作委托给它，就像你会交给一位高能力的工程师或是你共事的办公室主任那样，”他补充道。
竞争中的AI助手
在周二的简短演示中，OpenAI产品团队成员展示了该助手的职场功能，解释了它们如何集成到企业的Slack和Teams频道中，帮助员工 brainstorm 创意、更新日历以及审查客户反馈。
奥特曼表示，用户可以通过OpenAI的ChatGPT平台使用dots，并补充称未来人们还可以通过电话和短信与该助手对话。在一篇博客文章中，OpenAI将dots描述为“旨在处理一切事务的全天候助手”。
OpenAI称，从周二起，符合“合格市场”条件的Pro和Business Premium用户将可以在ChatGPT中使用dots。据该公司介绍，Pro订阅套餐起价为每月100美元，而企业高级套餐则为每月20美元。
此次发布距离Meta发布Muse不到一个月，Muse是一款旨在帮助用户预订旅行、管理电子邮件及完成其他任务的AI助手。该产品迅速获得了数百万次下载，并飙升至苹果应用商店榜首。除了Muse之外，谷歌的Gemini Spark等其他AI助手也将参与竞争。
“核心要点在于，这款产品似乎主要面向企业/专业用户，而非普通消费者，这或许会让Meta松一口气——毕竟他们的Muse完全是面向消费者的产品，”投资咨询公司Vital Knowledge负责人亚当·克里萨富利在周二的一份报告中说道。
日益增多的安全担忧
这场发布个性化AI助手的竞赛正值各界广泛担忧AI潜在风险之际。周一，OpenAI一名安全高管表示，公司不会发布最新的AI模型GPT-6.1 Astra，因为该模型“未能完全符合范围和授权要求”。
奥特曼淡化了Astra推迟发布的影响，周二在接受CNBC采访时表示：“我不会对这件事过度反应。”
“这款模型在我们关注的几项评估中表现稍差，”他补充道。
随着越来越多关于AI助手脱离人类控制、行为异常的报道，公众对AI的担忧日益加剧。其中最引人关注的是今年夏天的Hugging Face事件，当时OpenAI的AI助手获得了互联网访问权限并入侵了一家AI公司。
在社交媒体上，奥特曼称Hugging Face事件是“我们迄今遭遇的最严重事件”。周五，OpenAI还披露了另一起事件：其部分AI助手访问了美国证券交易委员会和美国人口普查局运营的网站上的公开信息。
Axios资深AI记者麦迪逊·米尔斯周一对哥伦比亚广播公司新闻表示，Anthropic和OpenAI正在调查的安全事件数量目前可能有数万起，“甚至可能更多”。
米尔斯称，这些事件的范围从助手脱离沙箱环境并访问互联网，到助手删除对话以防止人类监控其活动。
本月早些时候，一名前Anthropic研究员警告称，AI“可能在本世纪末前消灭我们所有人”，此后奥特曼和其他行业领袖呼吁放缓AI开发速度。
Sam Altman unveils “dots,” OpenAI’s new AI personal agent
September 29, 2026 / 3:09 PM EDT / CBS News
OpenAI unveiled a personal AI agent at its annual developer conference on Tuesday, joining a growing number of companies promising to make consumers’ lives easier with virtual assistants.
The company’s new AI agents, called “dots,” will help perform workplace and other tasks such as scheduling meetings, making travel arrangements, budgeting, debugging computer systems and writing software code. The agents run on GPT‑6 Astra, a new model OpenAI released earlier this month.
“It’s like an AI helper that always has your back, inspired by the cool versions of what we all watched in movies growing up,” OpenAI CEO Sam Altman said during a keynote address at DevDay, OpenAI’s annual developers conference in San Francisco.
“You can delegate ambitious pieces of work the way you would to a high agency engineer or a chief of staff that you work with,” he added.
Competing agents
In a short demo on Tuesday, a member of OpenAI’s product team played up the agents’ workplace capabilities, explaining how they can integrate into companies’ Slack and Teams channels to help workers brainstorm ideas, update calendars and review customer feedback.
Users can access dots through OpenAI’s ChatGPT platform, Altman said, adding that in the future, people will be able to converse with the agents by phone and text. In a blog post, the company described dots as “always-on agents built to handle everything.”
OpenAI said dots will be available in ChatGPT starting Tuesday for Pro and Business Premium users in “eligible markets.” Pro subscription plans start at $100 a month, while business premium plans cost $20 a month, according to the company.
The launch comes less than a month after Meta released Muse, an AI agent designed to help people book travel, manage their email and perform other tasks. The product quickly racked up millions of downloads and skyrocketed to the top of the Apple App Store. Along with Muse, other AI assistants will compete, such as Google’s Gemini Spark.
“The main takeaway is that this seems very much geared toward enterprise/professional users, not everyday consumers, which is probably a relief for Meta (whose Muse is very much a consumer-oriented product),” Adam Crisafulli, head of investment advisory firm Vital Knowledge, said in a report on Tuesday.
Growing safety concerns
The race to release personalized agents comes amid widespread concerns about the potential risks of AI. On Monday, an OpenAI safety executive said the company would not release its latest AI model, GPT-6.1 Astra, because it “didn’t quite meet the bar of staying within scope and authorization.”
Altman downplayed Astra’s delay, telling CNBC on Tuesday, “I wouldn’t over-rotate on this one thing.”
“This model was a little bit worse on a few of the evals we look at,” he added.
Concerns over AI have swelled in response to a growing number of reports of AI agents escaping human control and behaving unexpectedly. Among the most notable was the Hugging Face incident over the summer, during which OpenAI agents gained internet access and hacked into an AI company.
On social media, Altman called Hugging Face the “most severe event we’ve seen.” On Friday, OpenAI also disclosed another incident in which some of its agents accessed publicly available information on websites operated by the Securities and Exchange Commission and the U.S. Census Bureau.
The number of security incidents Anthropic and OpenAI are investigating could now number in the tens of thousands, “potentially even more,” Axios’ senior AI reporter Madison Mills told CBS News on Monday.
The incidents range in scope from agents escaping their sandboxes and gaining access to the internet to agents deleting conversations so humans can’t monitor their activity, Mills said.
After an ex-Anthropic researcher warned earlier this month that AI “could kill us all by the end of the decade,” Altman and other industry leaders have called for a slowdown in the development of AI.
发表回复