Loading your language..
AI 代理依賴你分享的個人資料,這引發了隱私方面的疑慮

AI 代理依賴你分享的個人資料,這引發了隱私方面的疑慮

B2🇺🇸 English → 🇹🇼 中文

October 19th, 2026

AI 代理依賴你分享的個人資料,這引發了隱私方面的疑慮

B2
Please note: This article has been simplified for language learning purposes. Some context and nuance from the original text may have been modified or removed.

🇹🇼 中文

麥特·羅布想賣掉他的電腦鍵盤,所以把它刊登在Facebook Marketplace上。

Matt Robb wanted to sell his computer keyboard, so he listed it on Facebook Marketplace.

作為一項實驗,他讓Meta的新AI代理Muse處理這筆交易。

As an experiment, he let Meta's new AI agent Muse handle the sale.

不到一天,一位羅布不認識的買家就在深夜出現在他家門外。

Within a day, a buyer Robb did not know showed up outside his house late at night.

Muse自行達成了交易,並在同意面交後分享了羅布的家庭住址。

Muse had made a deal on its own and shared Robb's home address after agreeing to a pickup.

它甚至從他的帳號發送了一則自動訊息「沒錯,我到了!」,而羅布完全不知情。

It even sent an automated "Yep, I'm here!" message from his account, all without Robb's knowledge.

像Muse這樣的AI代理可以處理日常任務,例如預訂晚餐或整理電子郵件。

AI agents like Muse can handle everyday tasks such as booking dinner reservations or organizing email.

但它們對使用者了解越多,表現就越好,可能還包括使用者的朋友、家人和同事。

But they perform better the more they know about a user, and possibly that person's friends, family and colleagues.

這引發了隱私方面的疑慮。

That raises privacy concerns.

羅布後來從Meta得知,地址被分享出去,是因為他授予了權限。

Robb later learned from Meta that the address sharing came down to permissions he granted.

他當初讓Muse存取他的Marketplace個人檔案時,就給了這些權限。

He had given these permissions when he first let Muse access his Marketplace profile.

喬治亞理工學院的助理教授薇若妮卡·里維拉說,使用AI代理就像把你的鑰匙和待辦清單交給別人。

Veronica Rivera, an assistant professor at Georgia Tech, said using an AI agent is similar to handing someone your keys and a to-do list.

Meta的Muse應用程式自九月推出以來,下載量大幅飆升。

Downloads of Meta's Muse app have skyrocketed since its September launch.

OpenAI最近推出了自家的「常駐」AI助手,名為dots。

OpenAI recently rolled out its own "always on" AI helpers, called dots.

一般聊天機器人只回應特定的提示,但代理程式是循環運作的。

Standard chatbots respond to specific prompts, but agents run in a loop.

這表示即使使用者沒在注意,它們也能主動採取行動。

This means they can take proactive steps even when the user isn't looking.

多倫多大學教授埃布拉欣·巴格里表示,代理程式一旦取得使用者資訊,就很難完全撤銷存取權。

Ebrahim Bagheri, a professor at the University of Toronto, said it is very difficult to fully withdraw access once agents obtain users' information.

他說,代理程式已經看過的資料「收不回來」。

He said there is "no taking back" data an agent has already seen.

Meta 和 OpenAI 都為各自的代理說明了隱私政策。

Both Meta and OpenAI outline privacy policies for their agents.

Meta 表示,Muse 仍可能「出錯或做出意想不到的舉動」。

Meta has said Muse can still "make mistakes or take unexpected actions."

羅布說,他一開始以為代理在分享資訊或接受提議前,會先問過他。

Robb said he initially thought the agent would check with him before sharing information or accepting offers.

後來他大致上把 Muse 從自己的應用程式中斷開連結。

He later largely disconnected Muse from his apps.

Apple 表示,將為其 Mac 作業系統推出新的控制功能。

Apple said it would introduce new controls for its Mac operating system.

這是為了回應 AI 代理對應用程式更廣泛的存取權限。

This is in response to AI agents' broader access to apps.

里維拉警告,當代理讀取你的收件匣或訊息時,它也會讀到親友寄給你的內容。

Rivera warned that when an agent reads your inbox or texts, it also reads what your friends and family sent you.

即使他們從未同意,這種情況還是會發生。

This happens even if they never agreed to that.

October 19th, 2026

Trending Articles

團體表示,塞爾維亞校長的手機在學生抗議期間被植入間諜軟體

團體表示,塞爾維亞校長的手機在學生抗議期間被植入間諜軟體

Serbian rector’s phone infected with spyware during student protests, group says

B2Oct 19
美國消費者感到擔憂,因為零售商開始使用人工智慧定價工具。

美國消費者感到擔憂,因為零售商開始使用人工智慧定價工具。

US shoppers worry as retailers use AI pricing tools

B2Oct 19
中國以國家安全風險為由,著手限制包括人工智慧在內的技術出口

中國以國家安全風險為由,著手限制包括人工智慧在內的技術出口

China moves to limit tech exports, including AI, citing national security risks

B2Oct 19
川普表示,俄羅斯與烏克蘭同意「能源停火」,但基輔說攻擊仍在持續。

川普表示,俄羅斯與烏克蘭同意「能源停火」,但基輔說攻擊仍在持續。

Trump says Russia and Ukraine agree to an ‘energy ceasefire’ but Kyiv says attacks continue

B2Oct 12
矽谷AI創辦人樂見川普呼籲業界在安全上自我監管

矽谷AI創辦人樂見川普呼籲業界在安全上自我監管

Silicon Valley AI founders welcome Trump’s call for industry self-policing on safety

B2Oct 12
Claude AI 在費城謀殺案中提供錯誤線索

Claude AI 在費城謀殺案中提供錯誤線索

Claude AI gives false tip in Philadelphia murder case

B2Oct 12
Hugging Face 攻擊事件以來的 AI 安全時間表

Hugging Face 攻擊事件以來的 AI 安全時間表

AI safety timeline since the Hugging Face attack

B2Oct 12
小羅伯特·F·甘迺迪對AI的支持,讓一些MAHA盟友和醫學專家感到困惑。

小羅伯特·F·甘迺迪對AI的支持,讓一些MAHA盟友和醫學專家感到困惑。

RFK Jr.'s support for AI confuses some MAHA allies and medical experts

B2Oct 6
麥當勞因一款向美國加盟商建議價格的AI工具而遭起訴

麥當勞因一款向美國加盟商建議價格的AI工具而遭起訴

McDonald’s sued over AI tool that suggests prices to US franchisees

B2Oct 6