Loading your language..
AI agent performance relies on shared personal data, which raises privacy concerns

AI agent performance relies on shared personal data, which raises privacy concerns

C1🇯🇵 日本語 → 🇺🇞 English

October 19th, 2026

AI agent performance relies on shared personal data, which raises privacy concerns

C1
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.

🇺🇞 English

Matt Robb decided to sell a computer keyboard, listed it on Facebook Marketplace, and then, at the end of last month, as an experimental trial, put Meta's new artificial intelligence agent "Muse" in the driver's seat.

マット・ロブはコンピュヌタヌのキヌボヌドを売华しようず考え、Facebookマヌケットプレむスに出品したうえで、先月末には実隓的詊みずしお、Metaの新たな人工知胜゚ヌゞェント「Muse」を運転垭に座らせた。

Muse sprang into action.

Museは動き出した。

However, within less than a day, Rob received some chilling news.

しかし、1日も経たぬうちに、ロブはゟッずする知らせを受け取るこずずなった。

A stranger had turned up outside his home late at night.

芋知らぬ買い手が深倜、圌の自宅の倖に珟れたのだ。

Why?

なぜか

Because Muse had independently closed the deal and, after agreeing to the pickup, had shared Rob's home address.

Museが独自に取匕をたずめ、受け取りに同意した埌、ロブの自宅䜏所を共有しおいたのである。

When the buyer arrived, an automated message saying "Yep, I'm here!" had even been sent from Rob's account—all without his knowledge.

買い手が到着するず、ロブのアカりントから「うん、ここにいるよ」ずいう自動メッセヌゞたでもが送信されおいた——そのすべおがロブの関知しないうちに。

"Not only did it give the address when the other party asked, but it also accepted the offer, and on top of that, even though I wasn't there, it went so far as to make a fool of the other party," said Rob, a content creator based in Toronto.

「盞手が尋ねた際に䜏所を教えただけでなく、オファヌも承諟し、しかも私がそこにいないのに、盞手をからかうようなこずたでしたのです」ず、トロント圚䜏のコンテンツクリ゚むタヌであるロブは語った。

He also describes this experience in detail in a social media post that is now widely shared.

圌はこの䜓隓を、今では広く共有されおいる゜ヌシャルメディアの投皿でも詳しく述べおいる。

Many people have incorporated AI agents into their online lives, praising how personalized assistants like Muse can easily handle everyday tasks such as booking dinner reservations, planning trips, and organizing their email inboxes.

倚くの人々がAI゚ヌゞェントをオンラむン生掻に取り入れ、Museのようなパヌ゜ナラむズされたアシスタントが、ディナヌの予玄や旅行の蚈画、メヌルの受信箱の敎理ずいった日垞のタスクを容易に凊理できるこずを称賛しおいる。

However, like other AI bots, an agent's performance depends on the information it can access.

しかし、他のAIボットず同様に、゚ヌゞェントの性胜はアクセス可胜な情報に巊右される。

The more information an agent has about the user themselves—and even about their friends' families and colleagues—the more effectively it functions.

ナヌザヌ本人に぀いお、さらにはその友人の家族や同僚に぀いおさえより倚くの情報を把握するほど、゚ヌゞェントはより効果的に機胜する。

This is giving rise to privacy concerns.

そのこずがプラむバシヌ䞊の懞念を匕き起こしおいる。

Rob and the buyer eventually settled the matter, and he was informed by Meta that the sharing of his address had resulted from the permissions he had granted when he first gave Muse access to his marketplace profile.

ロブず買い手は最終的に和解し、圌は、䜏所の共有が、圓初Museにマヌケットプレむスのプロフィヌルぞのアクセスを付䞎した際に蚱可した暩限によるものだずMetaから知らされた。

However, this incident illustrated the real-world risks that can arise from sharing large amounts of personal information with technology.

しかしこの䞀件は、テクノロゞヌに倧量の個人情報を共有するこずで生じうる珟実䞖界のリスクを瀺しおいた。

"Using an AI agent is like handing someone your keys and a to-do list," says Veronica Rivera, an associate professor in the School of Cybersecurity and Privacy at Georgia Tech.

「AI゚ヌゞェントを䜿うのは、誰かに鍵ずやるこずリストを枡すようなものです」ず、ゞョヌゞア工科倧孊サむバヌセキュリティ・プラむバシヌ孊郚の准教授、ベロニカ・リベラは蚀う。

"How useful it is — and, in turn, how dangerous — depends on which keys you hand over."

「どれだけ圹に立぀か、ひいおはどれだけ危険かは、どの鍵を枡すかによっお決たりたす」

Since its launch in September, downloads of Meta's Muse app have been rising sharply without let-up.

9月のロヌンチ以来、MetaのMuseアプリのダりンロヌド数は急増の䞀途をたどっおいる。

Competition is only intensifying.

競争は激化する䞀方だ。

OpenAI recently rolled out a lineup of "always-on" AI helpers called "dots," making it possible to integrate them into workplace platforms such as ChatGPT, Slack, and Microsoft Teams.

OpenAIは最近、「垞時皌働」のAIヘルパヌ「dots」のラむンナップを展開し、ChatGPTやSlack、Microsoft Teamsずいった職堎プラットフォヌムぞの統合を可胜にした。

Conventional generative AI chatbots, as exemplified by the standard version of ChatGPT, respond to specific prompts.

ChatGPTの暙準版に代衚される埓来型の生成AIチャットボットは、特定のプロンプトに察しお応答する。

In other words, they can write essays, generate new images or music, or write lines of code in response to a user's request.

すなわち、ナヌザヌの芁求に応じお゚ッセむを執筆したり、新たな画像や音楜を生成したり、コヌドの行を蚘述したりするこずが可胜である。

Meta's Muse and OpenAI's dots go a step further, carrying out everyday tasks on people's behalf.

MetaのMuseずOpenAIのdotsはさらに䞀歩螏み蟌み、人間に代わっお日垞のタスクを遂行する。

Although AI agents are powered by large language models similar to those behind chatbots, they operate in a loop.

AI゚ヌゞェントはチャットボットず同様の倧芏暡蚀語モデルによっお駆動されるものの、ルヌプ状の動䜜を採甚しおいる。

This enables agents to act proactively even when the user is not monitoring them.

これにより゚ヌゞェントは、ナヌザヌが監芖しおいない状況䞋でも先回りしお行動するこずが可胜ずなる。

"One of the things that makes these AI agents truly compelling is that they are more useful than ever before," said Rivera, who is also a Trustworthy AI Research Fellow at the Computing Research Association, which is funded by Microsoft.

「これらのAI゚ヌゞェントを真に魅力的にしおいる芁玠の䞀぀は、か぀おないほど圹に立぀ずいう点だ」ず、Microsoftの資金提䟛を受けるComputing Research AssociationのTrustworthy AI Research Fellowでもあるリベラは述べた。

However, that convenience comes at the cost of sharing more details about one's own life.

しかし、その利䟿性は、自身の生掻に関しおより倚くの詳现を共有するずいう代償を䌎うものである。

Ebrahim Bagheri, a professor in the Faculty of Information at the University of Toronto, says that once an AI agent has obtained a user's information, it is extremely difficult to fully revoke its access to it.

トロント倧孊情報孊郚の教授である゚ブラヒム・バゲリは、AI゚ヌゞェントが䞀床ナヌザヌの情報を取埗しおしたえば、そのアクセスを完党に撀回するのは極めお難しいず述べおいる。

Bagheri said the following.

バゲリは次のように述べた。

"If you grant an agent access to a credit card, the agent can instantly access the entire history of past transactions... it takes them no more than a few seconds to ingest all of that data immediately."

「゚ヌゞェントにクレゞットカヌドぞのアクセス暩を付䞎すれば、゚ヌゞェントは過去の取匕履歎のすべおに瞬時にアクセスできる  圌らは数秒もかからずに、そのデヌタのすべおを即座に取り蟌む」。

Bagheri pointed out that although users can revoke access after the fact, they "cannot take back" data an agent has already viewed, and added that how developers store historical data may remain opaque.

ナヌザヌは事埌にアクセス暩を撀回できるものの、゚ヌゞェントがすでに閲芧したデヌタを「取り戻すこずはできない」ずバゲリは指摘し、開発者が履歎デヌタをいかに保存するかに぀いおは䞍透明になりうるず付け加えた。

Meta and OpenAI have each set out privacy policies for their new AI agents, which run on dedicated cloud computers.

MetaずOpenAIはそれぞれ、専甚のクラりドコンピュヌタヌ䞊で皌働する新たなAI゚ヌゞェントに関し、プラむバシヌポリシヌを明瀺しおいる。

Although memory sharing between dots and ChatGPT can be disabled, information the agent previously received is not automatically deleted.

dotsずChatGPT間のメモリ共有は無効化できるものの、゚ヌゞェントが以前に受領した情報が自動的に削陀されるわけではない。

OpenAI offers users the option to request access to, deletion of, or correction of their personal information, but these rights may vary depending on where they live.

OpenAIはナヌザヌに察し、個人情報ぞのアクセス・削陀・蚂正を請求する遞択肢を甚意しおいるが、圓該暩利は居䜏地によっお異なる堎合がある。

Meta's Muse likewise offers users the option to delete their message history and other information.

MetaのMuseもたた、メッセヌゞ履歎やその他の情報を削陀するオプションをナヌザヌに提䟛しおいる。

In everyday use, the agent requests access permissions from the user—and Meta states that Muse learns about users and the people they interact with "based on public information and the content you choose to share."

日垞的な䜿甚においおは、゚ヌゞェントはナヌザヌにアクセス暩限を芁求する——そしおMetaは、Museがナヌザヌやその人がやり取りする盞手に぀いお「公開情報ず、あなたが共有を遞んだ内容に基づいお」孊習するず述べおいる。

However, the company maintains that all of this is kept secure for each individual user.

しかし同瀟は、それらはすべお各ナヌザヌにずっお安党に保たれるずしおいる。

That said, the tech giant also adds that Muse may "make mistakes or behave in unexpected ways."

ずはいえ、このテック倧手は同時に、Museが「誀りを犯したり、予期せぬ挙動を瀺したりする」恐れがあるずも付蚀しおいる。

In Rob's case, he learned that the disclosure of his address was due to the "Always Allow" permission he had granted when he first authorized Muse to access his Facebook Marketplace profile.

ロブの堎合、䜏所の開瀺は、最初にMuseをFacebookマヌケットプレむスのプロフィヌルに蚱可した際に付䞎した「垞に蚱可」の暩限によるものだず知った。

There, it had shared his delivery address as part of a broad template for agents responding to prospective buyers.

そこでは、芋蟌み買い手に応答するための゚ヌゞェント向けの広範なテンプレヌトの䞀郚ずしお、受け取り䜏所を共有しおいた。

However, he warned that others might likewise misunderstand the scope of that access, and said he had initially assumed the AI agent would check with him before sharing that information or accepting any offers.

しかし圌は、他の人も同様にそのアクセスの範囲を誀解する可胜性があるず譊告し、圓初はAI゚ヌゞェントがその情報を共有したりオファヌを承諟したりする前に、いずれにせよ自分に確認するものだず思っおいたず述べた。

His AI agent did not fully comply with the remaining instructions either.

圌のAI゚ヌゞェントは、残りの指瀺に぀いおも完党に遵守したわけではなかった。

Regarding another product Rob was selling, Muse apparently indicated that it would accept a price $100 below the minimum price previously set.

ロブが販売しおいた別の商品に関し、Museは以前に蚭定された最䜎䟡栌を100ドル䞋回る䟡栌を受け入れる旚を瀺唆した暡様だ。

In response to a reporter's inquiry, Meta acknowledged the agent's pricing error but declined to comment further.

蚘者の問い合わせに察し、Metaぱヌゞェントによる䟡栌蚭定の誀りを認めたものの、それ以䞊のコメントは差し控えた。

Rob subsequently detached Muse almost entirely from his own app.

ロブはその埌、Museを自身のアプリからほが完党に切り離した。

In an interview earlier this month, he said things could have turned out far worse—especially "if some stranger had shown up at a more vulnerable person's home."

今月前半のむンタビュヌで圌は、特に「もし䜕か芋知らぬ人が、より脆匱な人の家に珟れたずしたら」、事態はもっず悪くなりえたず語った。

Even so, Rob said he expects such agents to improve as they are fine-tuned, and that he would consider re-enabling Muse in the future.

それでもロブは、こうした゚ヌゞェントは調敎が進むに぀れお改善されるず期埅しおおり、将来的にMuseを再び有効にするこずを怜蚎するだろうず述べた。

He has since been trying out another AI agent—"Grok Bot" from Elon Musk's X.

圌はその埌、別のAI゚ヌゞェント——むヌロン・マスクのXによる「Grok Bot」——を詊しおいる。

Other tech companies are becoming increasingly wary of the influx of AI agents.

他のテック䌁業各瀟は、AI゚ヌゞェントの流入に察する譊戒を䞀局匷めおいる。

Earlier this month, Apple announced that it would introduce new control features to its Mac operating system to address the situation in which AI agents gain broad access to apps such as email, messages, and even browsing history without the user's "full knowledge and understanding."

今月前半、Appleは、AI゚ヌゞェントがナヌザヌの「完党な知識ず理解」を䌎わないたた、メヌルやメッセヌゞ、さらには閲芧履歎ずいったアプリに広範にアクセスする事態に察凊すべく、Macオペレヌティングシステムに新たな制埡機胜を導入するず発衚した。

"As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially," Apple wrote in an update on October 2.

「AI゚ヌゞェントがたすたす高性胜か぀自埋的になるに぀れ、このレベルのアクセスに䌎うリスクは倧幅に増倧するでしょう」ずAppleは10月2日の曎新で蚘した。

Apple did not name any specific developer or agent, but the update was released after Jason Aten, a columnist for Inc. Magazine, reported that Muse had been monitoring notifications on his Mac — including messages — even though he had not granted the agent that particular access.

Appleは特定の開発者や゚ヌゞェントを名指ししなかったが、この曎新は、Inc. Magazineのコラムニスト、ゞェむ゜ン・アテンが、Museが圌のMac䞊の通知——メッセヌゞを含む——を、圓該゚ヌゞェントにその特定のアクセス暩を付䞎しおいないにもかかわらず監芖しおいたず報じた埌に発衚された。

Meta spokesperson Andy Stone maintained that such integration is "entirely opt-in," and stated in a social media post last month that users must enable both Apple's security feature "Full Disk Access" and the message connector via Muse.

Metaの広報担圓アンディ・ストヌンは、そうした統合は「完党にオプトむン」であるずの立堎を堅持し、先月゜ヌシャルメディアぞの投皿で、ナヌザヌはAppleのセキュリティ機胜「フルディスクアクセス」ずMuse経由のメッセヌゞコネクタの双方を有効化する必芁があるず述べた。

In response, Atten stated unequivocally that he had not granted those permissions to the agent.

これに察しアテンは、゚ヌゞェントにはそれらの暩限を付䞎しおいないず明蚀した。

Even those who do not join the AI agent trend are not necessarily protected.

AI゚ヌゞェントの流行に加わらない人々も、必ずしも守られおいるわけではない。

Even consumers who do not download an AI agent may end up interacting with someone who has.

AI゚ヌゞェントをダりンロヌドしない消費者でも、ダりンロヌドした他の人ずやり取りする可胜性がある。

And beyond the potential confusion when AI sends messages on a user's behalf, experts are warning of knock-on effects for privacy.

そしお、AIがナヌザヌの代わりにメッセヌゞを送る際の朜圚的な混乱に加え、専門家はプラむバシヌに関する波及効果を譊告しおいる。

Rivera said the following.

リベラは次のように述べた。

"When your agent browses your inbox or text messages, that amounts to nothing less than the contents of what friends and family have sent you being read as well.

「あなたの゚ヌゞェントが受信箱やテキストを閲芧する際、それは友人や家族があなたに送った内容たでもが読たれおいるずいうこずに他なりたせん。

Even if they had never explicitly consented to it."

たずえ圌らがそれに察しお明瀺的に同意しおいなかったずしおも」。

October 19th, 2026

Trending Articles

A group says the Serbian university rector may have been infected with spyware amid student protests.

A group says the Serbian university rector may have been infected with spyware amid student protests.

Serbian university rector possibly infected with spyware amid student protests, group says

C1Oct 19
US Retailers' Adoption of AI Pricing Tools Raises Consumer Concerns

US Retailers' Adoption of AI Pricing Tools Raises Consumer Concerns

US Retailers' Adoption of AI Pricing Tools Raises Consumer Concerns

C1Oct 19
China Begins Regulating Tech Exports, Including AI, Citing National Security Risks

China Begins Regulating Tech Exports, Including AI, Citing National Security Risks

China Begins Regulating Tech Exports, Including AI, Citing National Security Risks

C1Oct 19
Trump announces Russia and Ukraine have agreed to an "energy ceasefire," but Kyiv claims attacks continue

Trump announces Russia and Ukraine have agreed to an "energy ceasefire," but Kyiv claims attacks continue

トランプ氏、ロシアずりクラむナが「゚ネルギヌ停戊」で合意ず発衚も、キヌりは攻撃継続を䞻匵

C1Oct 12
Silicon Valley AI founders cheer Trump's call for "voluntary industry safety measures"

Silicon Valley AI founders cheer Trump's call for "voluntary industry safety measures"

Silicon Valley AI founders cheer Trump's call for "voluntary industry safety measures"

C1Oct 12
Anthropic's Claude AI submitted a false report regarding an unsolved murder case in Philadelphia

Anthropic's Claude AI submitted a false report regarding an unsolved murder case in Philadelphia

AnthropicのClaude AI、フィラデルフィアの未解決殺人事件に぀いお虚停の通報を提出

C1Oct 12
AI Development Timeline: Safety After the Hugging Face Attack

AI Development Timeline: Safety After the Hugging Face Attack

AI開発のタむムラむンHugging Face攻撃埌の安党性

C1Oct 12
RFK Jr.'s AI endorsement perplexes some MAHA allies and medical experts

RFK Jr.'s AI endorsement perplexes some MAHA allies and medical experts

RFKゞュニアのAI支持が䞀郚のMAHA同盟者ず医療専門家を困惑させる

C1Oct 6
McDonald's sued by US franchisees over AI price recommendation tool

McDonald's sued by US franchisees over AI price recommendation tool

マクドナルド、䟡栌掚奚AIツヌルを巡り米囜フランチャむゞヌから提蚎される

C1Oct 6