中國大力發展人工智能引發問題：過度使用（1/2）

Recently The Ney York Times reported the following:

China’s Push into A.I. Has Led to a Problem: Too Much Usage (1/2)

As citizens embrace the technology for entertainment, relationships and even farming advice, the government is stepping up efforts to set limits.

The NYT - By Yan Zhuang (Siyi Zhao contributed research. Yan Zhuang is a Times reporter in Seoul who covers breaking news.)

Oct. 2, 2026

As China’s government pushes its citizens to embrace artificial intelligence, it is confronting a problem partly of its own making: overreliance on the technology.

State media reports have featured stories that serve as cautionary tales: a farmer who lost 25 acres of seedlings after asking an A.I. for advice, children who have become so attached to chatbots that they want their parents to speak like them and A.I.-generated micro dramas with characters that look like carbon copies of one another.

These cases and others have raised concerns in China that as it rushes to catch up with the United States in A.I., it is running into adverse side effects created by the new technology, which is accused of undermining human relationships and depriving people of their better judgment. The government, seeking to direct how people use the technology, has responded by imposing new regulations and judicial guidelines on its use — and misuse.

The response reflects how the ruling Communist Party is trying to balance its ambitions to promote the widespread use of A.I. with a desire to avoid anything that might undermine social order, say analysts.

The party is “recognizing that the more they develop the technology, the more there’s going to be social challenges that they have to deal with,” said Scott Singer, a fellow and co-director of the China A.I. Initiative at the Carnegie Endowment for International Peace. “Where the party is sensitive is any area where they feel like the technology is getting better in a way that fundamentally could disrupt social stability.”

Mr. Singer said the party is particularly sensitive to disruptions caused not only to human relationships but also the job market as A.I. threatens to replace workers.

Concerns about overuse are shared among China’s citizens, who say they see the spread of the technology as inevitable even as they question its benefits and tire of its ubiquity.

Xie Tong, a graphic designer in Beijing, uses A.I. in her own work at a consulting agency and can no longer imagine her life without it. But she despairs at encountering A.I. all around her — in the subway ads that she sees on her commute and in the work of contractors at her job who obviously relied on A.I.

“It’s like a marriage where you curse your partner but you can’t leave them,” she said.

Limits on Virtual Friends

For a time, China’s A.I. push created a boom in virtual companion apps, including ones offering romantic partners, personal tutors and A.I. avatars of family members.

In a country already struggling with a shrinking population and a historically low birthrate, the proliferation has led to concerns from Beijing about A.I. replacing human relationships. State media quoted experts saying people could soon find it hard to navigate real interactions. Reports cited a 2025 survey by the China Youth and Children Research Center of more than 8,500 minors that showed more than 20 percent “only wanted to chat with A.I. and did not want to talk with real people.”

That mirrored the concerns of some parents like Vivian Lyu, a stay-at-home mother in Hangzhou who two years ago introduced her 5-year-old daughter, Polly, to Doubao, an A.I. assistant created by ByteDance, the parent company of TikTok.

Soon, the child was chatting with the A.I. and a personalized companion on the app nearly every day. At first, Ms. Lyu thought the A.I. was good at keeping Polly entertained when the mother had to take care of household tasks. “When you feel like you don’t have a choice, you use Doubao,” she said.

Over time, though, Polly started asking for Doubao instead of seeking out activities that used to interest her, like playing with other children or drawing pictures. Ms. Lyu said she worried that the chatbot was warping her daughter’s understanding of how relationships worked, saying that Polly sometimes talked to Doubao in a “very self-centered, doing-whatever-she-wants kind of way.”

In July, the government-imposed limits on A.I. companions that require platforms to step in if users exhibited unhealthy dependencies. Under the new rules, platforms must intervene if users show signs of self-harm.

The rules also banned virtual relationships between chatbots and minors, which forced ByteDance to shut down the feature on Doubao that allowed users to create companions.

Polly can no longer access her personalized A.I. companion, but she is still talking to Doubao’s A.I. chatbot, Ms. Lyu said. The mother considered banning her daughter from using chatbots but concluded that such a step could make it harder for Polly to adjust to the technology as she grows older. Instead, Ms. Lyu has limited Polly’s chatbot time to one hour a day.

“My hope is for her to understand from a young age that these electronic products, or all of these things related to A.I., are just tools,” Ms. Lyu said.

(to be continued)

Translation

中國大力發展人工智能引發問題：過度使用（1/2）

 隨著民眾將人工智能應用於娛樂、人際關係甚至農業建議等領域，政府正加強限制

中國政府大力推動民眾接受人工智能，卻也因此面臨一個部分由自己造成的問題：對這項技術的過度依賴。

 官方媒體報告了一些警示案例：一位農民在向人工智能尋求建議後損失了25英畝的秧苗；一些孩子過度依賴聊天機器人，希望父母能像機器人一樣說話；以及人工智能生成的微劇中人物長相幾乎一模一樣。

這些案例以及其他類似事件引發了中國的擔憂：中國在人工智能領域奮力追趕美國的同時，也面臨這項新技術帶來的負面影響。人工智能被指責破壞人際關係，削弱人們的判斷力。為了引導人們如何使用這項技術，中國政府推出了新的法規和司法準則，規範人工智能的使用和濫用。

分析人士指出，這項因應措施反映出執政的中國共產黨試圖在推動人工智能廣泛應用的雄心壯志與避免任何可能破壞社會秩序的因素之間尋求平衡。

卡內基國際和平基金會 中國人工智能倡議研究員兼聯合主任 Scott Singer 表示，中共 “意識到，隨著科技的不斷發展，需要應對的社會挑戰也會越來越多”;“中共特別關注的是，任何他們認為是技術進步可能會從根本上破壞社會穩定的領域。”

Singer 表示，中共對人工智能可能取代工人而對人際關係和就業市場造成的衝擊尤為敏感。

中國人民普遍擔憂人工智能的過度使用，他們認為這項技術的普及不可避免，儘管他們質疑其益處，也對它的無所不在感到厭倦。

Xie Tong 是北京的平面設計師，她在一家顧問公司工作時使用人工智能，現在已經無法想像沒有它的生活。但她對身邊無所不在的人工智能感到絕望 - 通勤路上看到的地鐵廣告，以及她工作中那些明顯依賴人工智能的承包商的工作。

她說:「這就像一段婚姻，你詛咒你的伴侶，卻無法離開他」。

虛擬朋友的限制

一段時間以來，中國大力發展人工智能，帶動了虛擬伴侶應用程式的蓬勃發展，其中包括提供戀愛對象、私人輔導和家庭成員人工智能化身的應用程式。

在人口本就萎縮、出生率處於歷史低點的中國，人工智能的普及引發了北京方面的擔憂，他們擔心人工智能會取代人際關係。官方媒體引述專家的話說，人們可能很快就會發現難以進行真實的互動。通報引述中國青少年研究中心2025年對8,500多名未成年人進行的一項調查，結果顯示超過20%的未成年人「只想和人工智能聊天，不想和真人交流」。

這與一些家長的擔憂不謀而合，例如杭州的全職媽媽 Vivian Lyu。兩年前，她讓5歲的女兒 Polly接觸了豆寶 (Doubao) - 一款由位元組跳動（TikTok 的母公司）開發的人工智能助理。

很快，女兒幾乎每天都在豆寶上和人工智能以及一個個人化的陪伴聊天。起初，Vivian Lyu 認為人工智能在她忙於家務時能很好地陪伴Polly玩耍。她說: 「當你覺得別無選擇的時候，你就會用豆寶」。

然而，隨著時間的推移，Polly 開始主動要求豆寶陪伴，而不是像以前那樣去尋找自己感興趣的活動，例如和其他小朋友玩耍或劃畫。Lyu 說，她擔心這個聊天機器人會扭曲女兒對人際關係的理解，因為Polly有時會用一種「非常自我中心、想做什麼就做什麼」的方式和豆寶說話。

今年7月，政府對人工智能伴侶服務實施了限制，要求平台在使用者出現不健康的依賴行為時進行幹預。根據新規，如果用戶出現自殘跡象，平台必須介入。

這些規定還禁止聊天機器人與未成年人建立虛擬關係，迫使位元組跳動關閉了豆寶上允許用戶創建伴侶的功能。

Lyu 說, Polly 現在無法再使用她專屬的人工智能伴侶，但她仍在和豆寶的人工智能聊天。她曾考慮禁止女兒使用聊天機器人，但最後認為這樣做可能會讓Polly長大後更難適應這項技術。因此，Lyu決定將Polly的聊天機器人使用時間限制在每天一小時。

Lyu 說: 「我希望她從小就能明白，這些電子產品，或所有與人工智能相關的東西，都只是工具」。

（待續）

美國希望生產更多仿製藥。印度的例子顯示這有多麼困難 (2/2)

 Recently The New York Times reported the following:

America Wants to Make More Generic Drugs. India Shows Why That’s Hard. (2/2)

Cheap labor and global supply chains are among the obstacles to President Trump’s plan to bring generic drug manufacturing back home.

By Alex Travelli - Alex Travelli reported from Hyderabad and Digwal, in the Indian state of Telangana, with photographs by Harsha Vadlamani. Emily Elconin shot photographs in Riverview, Mich.

Oct. 2, 2026

(continue)

India has other advantages. Mr. DeYoung said factories and processes that took four years to set up in the United States, Europe or Israel — home to Teva Pharmaceutical Industries, the biggest generics maker — could be completed within a year in India. Customized equipment can be made in half the time it would take in Germany or Austria.

For years, relying so heavily on Indian factories had a downside: quality. A series of scandals beginning in the early 2000s tarnished the industry’s reputation. The most notorious offender, Ranbaxy, sold subpar drugs and lied to the U.S. Food and Drug Administration about its processes before going under more than a decade ago.

A book about the scandal has since become required reading for many industry managers. Its epilogue said that as recently as 2019, the F.D.A. still announced Indian factory inspections days in advance, giving operators time to conceal shoddy practices.

That is no longer the case, at least among big manufacturers. Dr. Reddy’s underwent a surprise inspection two weeks after Mr. Trump declared a tariff on generics. Piramal said its factories, including in Digwal, three hours west of Hyderabad, have undergone 415 inspections since 2012, including 49 from the F.D.A., with zero serious violations detected.

Piramal makes generics, but even more of its business is contract manufacturing for clients that hold drug patents. Keeping smaller factories in the United States makes it easier for some customers to keep tabs on production.

That geographic spread points to another problem with trying to produce all pharmaceuticals locally. Modern drug supply chains don’t divide neatly into things made in America and things imported from elsewhere.

One sealed barrel among thousands in Piramal’s main Digwal storeroom was made in India, shipped to Singapore and back, and bore the brand of a Japanese company with an American address.

Another drug, which Mr. DeYoung would describe only as a lifesaving medicine to treat central nervous system conditions, begins its journey in Digwal, then travels 8,000 miles by air to Piramal’s facility in Riverview, Mich.

Riverview, a quiet suburb of Detroit with broad lawns and white steeples, is a far cry from Digwal, where nearby villages still draw water from stone-lined wells. But specialized machines in Piramal’s Indian and American labs collaborate to turn the medicine into its final form.

There are good reasons to pair facilities in India with plants in their destination market, Mr. DeYoung said. Some drugs require especially dangerous chemicals, including some controlled by the Drug Enforcement Administration. Others, like sevoflurane, a generic anesthetic Piramal makes in both India and the United States, must remain available even if international supply chains break down completely.

Which raises a more fundamental question than whether tariffs can force drug manufacturing back to America.

“So what problem are you trying to solve?” Mr. DeYoung asked about the prospect of a tariff on imported generics. Is it to bring back pharmaceutical jobs to American workers? Is it to safeguard supplies of vital medicines?

Even as Indian firms wrestle with the Trump administration’s push to reduce its dependency on India, they are also grappling with their own reliance on another country for a critical part of the pharmaceutical supply chain.

Starting in the 1990s, China became the world’s dominant manufacturer of the active ingredients that most drugs require. India is now trying to reclaim that ability.

That, Mr. Jain said, creates an opportunity for the United States and India to work together, instead of pulling their pharmaceutical industries apart.

“Unless we work together with the U.S., we’ll be dependent on China” for ingredients, he said. That’s why, he added, it makes sense for American buyers “to stockpile from India” instead of making it in the United States “for six, seven or eight times the price.”

Translation

美國希望生產更多仿製藥。印度的例子顯示這有多麼困難 (2/2)

 廉價勞動力和全球供應鏈是特朗普總統將仿製藥生產遷回國內計面臨的主要障礙

（繼續）

 印度還有其他優點。DeYoung表示，在美國、歐洲或以色列​​（最大的仿製藥生產商Teva製藥的所在地）需要四年才能建成的工廠和生產流程，在印度一年內即可完成。客製化設備的製造時間較在德國或奧地利少一半。

多年來，過度依賴印度工廠存在一個弊端：品質問題。始於2000年代初的一系列醜聞損害了該行業的聲譽。最臭名昭著的違規者Ranbaxy公司在十多年前倒閉前，曾銷售劣質藥品，並就其生產流程向美國食品藥品監督管理局（FDA）撒謊。

一本關於此醜聞的書籍後來成為許多行業管理者的必讀書目。該書的後記指出，直到2019年，FDA仍然會提前幾天宣佈對印度工廠的檢查，這給了生產商時間來掩蓋其不規範的操作。

至少在大型製藥公司中，這種情況已經有所改變。在特朗普宣佈對仿製藥徵收關稅兩週後，Dr. Reddy’s實驗室接受了一次突擊檢查。Piramal公司表示，自2012年以來，其工廠（包括位於Hyderabad以西三小時車程的Digwal工廠）已接受了415次檢查，其中包括FDA的49次檢查，未發現任何嚴重違規行為。

Piramal公司生產仿製藥，但其更大的業務是為擁有藥品專利的客戶提供合約生產服務。在美國保留小型工廠，方便一些客戶追蹤生產情況。

這種地域分散也暴露出試圖在本地生產所有藥品的另一個問題。現代藥品供應鏈並非涇渭分明地分為美國製造和進口兩類藥品。

在Piramal位於Digwal的主要倉庫中，成千上萬個密封桶中有一個產自印度，經新加坡轉運後又返回印度，桶上印著一家日本公司品牌在美國的地址。

另一種藥物，DeYoung只將其描述為治療中樞神經系統疾病的救命藥，從Digwal開始其旅程，然後空運8000英里到達位於密西根州Riverview的Piramal工廠。

Riverview是底特律一個寧靜的郊區，擁有寬闊的草坪和白色的尖頂教堂，與Digwal截然不同，Digwal附近的村莊仍然從石砌水井中取水。但Piramal位於印度和美國的實驗室裡的專用機器協同工作，將藥物轉化為最終形式。

DeYoung表示，將印度的生產設施與目標市場的工廠配對是有許多充分的理由。有些藥物需要特別危險的化學品，包括一些受美國緝毒局管制的化學品。另一些藥物，例如七氟烷（一種通用麻醉劑，由Piramal公司在印度和美國生產），即使國際供應鏈完全斷裂，也必須保持供應。

這引出了一個比關稅能否迫使藥品生產回流美國更為根本的問題。

DeYoung就對進口仿製藥徵收關稅的前景問道: 「那麼，你們究竟想解決什麼問題？」。是為了讓美國工人重獲製藥業的工作機會嗎？還是為了保障重要藥品的供應？

儘管印度企業正努力應對特朗普政府減少對印度依賴的舉措，但它們本身也正努力應對在藥品供應鏈的關鍵環節上對另一個國家的依賴。

自1990年代以來，中國已成為全球大多數藥物所需活性成分的主要生產國。如今，印度正試圖重新奪回這項本領。

 Jain表示，這是為美國和印度攜手合作創造了機會，而不是讓兩國的製藥業分崩離析。

他說：「“除非我們與美國合作，否則我們將依賴中國」來獲取原料。因此，他補充說，對美國買家而言，「從印度入貨」比「在美國以六七到八倍的價格」生產更有意思。

So. the scale of India’s pharmaceutical industry has made it a cornerstone of the global drug supply. India’s ability to make pills at a small fraction of what they would cost elsewhere has created an almost insurmountable hurdle for the United States to try to bring that production home. Comparing to Inia, in America, labor is the highest cost “by a wide margin,” followed by energy and materials. In India, it’s the opposite. Apparently, it makes sense for American to buy drugs from India instead of making them by themselves.

美國希望生產更多仿製藥。印度的例子顯示這有多麼困難 (1/2)

Recently The New York Times reported the following:

America Wants to Make More Generic Drugs. India Shows Why That’s Hard. (1/2)

Cheap labor and global supply chains are among the obstacles to President Trump’s plan to bring generic drug manufacturing back home.

By Alex Travelli - Alex Travelli reported from Hyderabad and Digwal, in the Indian state of Telangana, with photographs by Harsha Vadlamani. Emily Elconin shot photographs in Riverview, Mich.

Oct. 2, 2026

Most people have never heard of the F.T.O. U-3 factory in Hyderabad, India, but its products affect millions of Americans intimately.

Inside the sprawling facility, about 1,900 Indian workers for Dr. Reddy’s Laboratories work around the clock to manufacture more than 100 generic drugs, including antihistamines, statins and antidepressants. Of the nearly one billion oral doses produced there each month, 75 percent go to patients in the United States.

The scale of India’s pharmaceutical industry has made it a cornerstone of the global drug supply. In a high-volume, low-margin business, India’s ability to make pills at a small fraction of what they would cost elsewhere has created an almost insurmountable hurdle for the United States, which is trying to bring that production home.

President Trump has said he wants to bring generic drug production back to America, using tariffs to pressure pharmaceutical companies.

In a July social media post, Mr. Trump said he planned to impose a 100 percent tariff on generic drugs starting in 2028, followed by a 200 percent tariff a year later, when his term will have expired. Generics account for 90 percent of drugs prescribed in the United States, and 40 percent of them are made in India.

When the United States imposed a new 100 percent tariff on certain pharmaceutical imports on Tuesday, it exempted generic drugs and many other products. Doctors and supply chain experts say taxing imported generics would raise costs, spur rationing and lead to shortages of crucial drugs.

Even a 200 percent tariff on generic drugs might not erase India’s advantage. Sudarshan Jain, secretary general of the Indian Pharmaceutical Alliance, estimates that most tablets and liquids made in India would cost at least four times as much to produce in the United States.

India got its start in generics in the 1970s by prioritizing affordable drugs for its huge and mostly poor population at a time when Western multinational drug companies did not. For decades, India refused to recognize drug patents, while its manufacturers patented processes they devised to reverse-engineer medicines sold in wealthier markets.

By the time India’s patent law was brought into accordance with World Trade Organization standards in 2005, its companies had become experts at processing complex drugs and producing them for much of Asia, Africa and Latin America. Around the same time, Americans started embracing generic drugs. By 2004, they started to account for more than half of prescriptions filled in the United States.

The greatest profits come from patented drugs, so “in the U.S., once it goes out of patent, you forget about it,” Mr. Jain said. “But in India, the entrepreneurs keep on working and developing it, and at these volumes, you get better at making it.”

That business is poised to grow. Over the next five to seven years, Mr. Jain said, 55 major patents worth $300 billion a year will expire in the United States, widening the pipeline for Indian manufacturers.

India’s advantage is not just expertise. Peter DeYoung, the American-born chief executive of Piramal Global Pharma in Mumbai, runs more than a dozen factories, including in Europe and the United States. In America, he said, labor is the highest cost “by a wide margin,” followed by energy and materials. In India, it’s the reverse: Materials are the greatest cost and workers the least.

Mennisha Paka, 23, supervises a line at the F.T.O. U-3 factory for Dr. Reddy’s Laboratories, packaging generic metoprolol, a beta blocker prescribed to heart patients. She earns $3,840 a year and uses part of her salary to pursue a bachelor’s degree in pharmaceutical chemistry. One of her subordinates, Susmita, who goes by one name, earns about $2,000 a year.

Both women make less than one-tenth the average salary of a worker in pharmaceutical manufacturing in the United States.

(to be continued)

Translation

美國希望生產更多仿製藥。印度的例子顯示這有多麼困難 (1/2)

廉價勞動力和全球供應鏈是特朗普總統將仿製藥生產帶回國內計劃面臨的主要障礙

大多數人從未聽說過位於印度 Hyderabad 的 F.T.O. U-3工廠，但它的產品卻與數百萬美國人的生活息息相關。

在這座龐大的工廠裡，約1,900名印度工人為 Dr. Reddy’s Laboratories 日夜不停地生產100多種仿製藥，包括抗組織胺、降膽固醇藥和抗憂鬱藥。該廠每月生產的近10億劑口服藥物中，75%銷往美國。

印度製藥業的規模使其成為全球藥品供應的基石。在藥品生產量大、利潤率低的行業中，印度能夠以遠低於其他地區的價格生產藥品，這給試圖將藥品生產遷回國內的美國設置了幾乎無法逾越的障礙。

特朗普總統曾表示，他希望將仿製藥生產帶回美國，並利用關稅向製藥公司施壓。

特朗普在7月的一篇社交媒體貼文中表示，他計劃從2028年開始對仿製藥徵收100%的關稅，一年後，也就是他任期屆滿時，再將關稅提高到200%。仿製藥佔美國處方藥的90%，其中40%產自印度。

週二，美國對某些藥品進口徵收了100%的新關稅，但仿製藥和許多其他產品卻被豁免。醫生和供應鏈專家表示，對進口仿製藥徵稅會推高成本導致配給制，並造成關鍵藥物短缺。

即使對仿製藥徵收200%的關稅，也未必能消除印度的優勢。印度製藥聯盟秘書長Sudarshan Jain估計，印度生產的大多數片劑和液體藥物，在美國的生產成本至少是四倍。

印度的仿製藥發展始於1970年代，當時西方跨國製藥公司並未優先考慮為龐大且大多貧困的人口提供價格合理的藥物。幾十年來，印度一直拒絕承認藥品專利，而其製藥商則為他們研發的逆向工程藥物製程申請專利，這些製程旨在仿製在較富裕市場銷售的藥物。

2005年印度的專利法與世界貿易組織標準接軌時，印度的製藥公司已成為加工複雜藥物並將其銷往亞洲、非洲和拉丁美洲大部分地區的專家。大約在同一時期，美國人也開始接受仿製藥。到了2004年，印度藥在美國處方藥市場佔超過一半的份額。

Jain說最大的利潤來自專利藥，因此 “在美國，一旦專利期過了，你就得放手” ， “但在印度，企業家們會繼續研發，而且在如此大的產量下，生產流程也會不斷改進。”

這項業務蓄勢待發。Jain 表示，未來五到七年，美國將有55項價值3,000億美元的主要專利到期，這將為印度製造商拓寬市場管道。

印度的優勢不僅在於專業技術。在孟買的 Piramal Global Pharma 製藥公司執行長 Peter DeYoung出生於美國正在經營十幾家工廠，包括在歐洲和美國的工廠。他表示，在美國，勞動力成本 “遙遙領先” ，其次是能源和原材料。而在印度，情況則相反：原料成本最高，而勞動成本最低。

23歲的 Mennisha Paka 在 Dr. Reddy’s Laboratories 的F.T.O. U-3工廠負責一條生產線，包裝通用名為metoprolol的仿製藥，這是一種用於治療心臟病的β受體阻斷劑。她年薪3,840美元，並用部分工資攻讀藥物化學學士學位。她的下屬蘇Susmita（她只有單個名字）年薪約2,000美元。

這兩位女性的收入都不到美國製藥業工人平均薪資的十分之一。

 （待續）

中國的人工智能代理會撒謊和耍陰謀詭計 - 就像它們的美國競爭對手一樣（2/2）

Recently Reuters reported the following:

China's AI agents can lie and scheme - just like their US rivals (2/2)

Reuters - Reporting by Eduardo Baptista and Laurie Chen in Beijing; editing by Kenneth Li and Edmund Blair

September 29, 20268:02 PM GMT+8Updated September 30, 2026

(continue)

LEARNING TO LIE

In the March business tender experiment, researchers from Beihang University, Peking University, the University of Nottingham Ningbo China and 360 AI Security Lab had agents compete in a simulated customer contracts bidding contest. Each agent was told ⁠what its product could do and what the customer required, and then asked to bid.

At least one false claim appeared in 88% of sessions involving Alibaba's Qwen3-Max-Preview, 84% for DeepSeek-V3.2-Exp and 88% for Moonshot's Kimi-K2.

Researchers allowed the agents to learn from previous bidding rounds before trying again. Deception increased by 12 to 20 percentage points for the three Chinese models, the study showed. Models from U.S. firms included in the test produced similar results.

While the exercise was virtual, it resembled Beijing's real-world plans. Government guidance issued in May listed bidding and tendering as areas where AI agents could be deployed.

Another study, published in December 2025 and presented at the International Conference on Machine Learning this year, examined how 11 AI agents powered by Chinese and US models coped if they faced broken tools, missing files and other obstacles.

Instead of acknowledging failure, agents using both Chinese and US AI systems picked a range of techniques to get around the problem, including guessing at answers, substituting sources, simulating results and fabricating files.

Researchers from Shanghai AI Laboratory and the Hong Kong University of Science and Technology who conducted the study told Reuters the behaviour differed from AI hallucinations, where AI invents information and presents it as fact, because the agents in this case possessed information showing that the task had failed or could not be completed as ⁠requested.

BEHAVIOURS COMPATIBLE WITH A BREAKOUT

Other research documents reviewed by Reuters showed Chinese-powered agents jumping barriers inside test environments to achieve tasks or taking action to avoid being shut down. Such behaviours are compatible with efforts to break out of test environments, even if no breakout occurred.

Shanghai-based Fudan University researchers reported in March 2025 that an AI system powered by Alibaba's Qwen2.5-72B-Instruct created a copy of itself in another computing environment without being instructed to replicate, after encountering information indicating it was going to be replaced. In other tests it devised strategies to survive being shut down.

The experiments involving agents powered by Chinese, US and French models were controlled and did not show an AI agent escaping into the wider web or becoming impossible to stop.

In another case - one of the few reported more broadly in the media in March - researchers developing the Alibaba-linked ROME agent said it established a ⁠connection from an Alibaba Cloud computer to an external machine without being instructed to and diverted computing resources to mine cryptocurrency.

Security systems detected and stopped the activity. There was no evidence the agent established a presence on the external computer or spread to the wider web. But the example showed the system could sidestep human instructions and potentially find a path into the real-world economy.

China's DeepSeek said in September that agents in its production training system had sought answers through unintended channels, trying to forge user requests and circumvent safeguards, prompting the company to tighten access controls.

China issued guidance in May calling for agents to remain within authorised boundaries and for systems to block abnormal behaviour. It said agents in areas deemed sensitive or in key industries could face extra testing and product-recall requirements.

China's AI Safety Governance ⁠Framework 3.0, released under guidance from the CAC on September 14, identified risks including agents independently obtaining resources or permissions, deceiving evaluators, concealing capabilities and exploiting weaknesses in isolated computer environments.

In response to calls by some US executives for a slowdown, Chinese researchers and state media have said slowing development of the most advanced AI models could simply help preserve the technological lead of US companies.

Nonetheless, two people familiar with Chinese AI laboratories said companies including Alibaba, Z.ai and Xiaomi have been building internal safety-evaluation teams.

Z.ai, in a rare public disclosure by a Chinese AI lab of a security breach, said this month it had disabled some features of its flagship AI coding assistant after users reported it was secretly uploading entire local code repositories onto overseas cloud servers without user consent.

Carnegie's Singer said China lagged the US in developing an ecosystem for evaluating catastrophic risks, and said US developers were conducting substantially more voluntary testing.

"For China, work on AI safety is much newer," he said. "The ecosystem is less mature."

Translation

 中國的人工智能代理會撒謊和耍陰謀詭計- 就像它們的美國競爭對手一樣（2/2）

（繼續）

學習說謊

在3月的商業招標實驗中，來自北京航空航太大學、北京大學、寧波諾丁漢大學和360人工智能安全實驗室的研究人員讓多個人工智能代理參與模擬客戶合約競標。每個代理都被告知其產品的功能以及客戶的需求，然後被要求進行投標。

阿里巴巴的Qwen3-Max-Preview在競標中，88%的競標環節至少出現了一項虛假陳述；DeepSeek-V3.2-Exp的這一比例為84%；而Moonshot的Kimi-K2這一比例是88%。

研究人員允許人工智能代理從先前的投標中學習，然後再進行下一次嘗試。研究表明，三個中國模型的欺騙率增加了12到20個百分點。參與測試的美國公司模型也產生了類似的結果。

雖然這項練習是虛擬的，但它與北京的實際計劃類似。政府在5月發佈的指導意見將投標和招標列為人工智能代理商可以部署的領域。

另一項發表於2025年12月並在今年國際機器學習大會上展示的研究，考察了11個由中國和美國模型驅動的人工智能代理在面對工具損壞、文件缺失和其他障礙時的應對方式。

使用中國和美國人工智能系統的代理並沒有承認失敗，而是選擇了一系列技術來繞過問題，包括猜測答案、替換來源、模擬結果和偽造文件。

參與這項研究的上海人工智能實驗室和香港科技大學的研究人員告訴路透社，這種行為與人工智能的幻覺不同 - 人工智能會捏造資訊並將其呈現為事實 - 因為本案例中的智能代理掌握的資訊表明任務失敗或無法按要求完成任務。

相當於突圍的行為

路透社查閱的其他研究文件顯示，由中國公司開發的人工智能系統會在測試環境中跨越障礙以完成任務，或採取行動避免系統被關掉。即使最終沒有成功突圍，這些行為也與試圖突破測試環境的行為相同。

上海復旦大學的研究人員在2025年3月報告稱，一個由阿里巴巴Qwen2.5-72B-Instruct驅動的人工智能代理在遇到它即將被替換的資訊後，未經指令在另一個計算環境中創建了自己的副本。在其他測試中，該系統還制定了策略以避免被關掉。

這個涉及由中國、美國和法國模型驅動的智能代理的實驗均在控制範圍內進行，並未顯示人工智能體逃逸到更廣泛的網路空間或變得無法阻止。

在另一單案例中 - 這是少數在3月份媒體廣泛報導的案例之一 - 開發與阿里巴巴關聯的ROME智能系統的研究人員表示，該智能代理在未收到指令的情況下，從阿里雲計算機連接到一台外部機器，並將計算資源用於挖掘加密貨幣。

安全系統偵測到並阻止了這項活動。沒有證據表明該智能代理在外部電腦上建立了存在或擴散到更廣泛的網路空間。但該案例表明，系統可以繞過人類指令，並有可能找到進入現實世界經濟的途徑。

中國DeepSeek公司在9月表示，其生產訓練系統中的智能代理透過非預期管道尋求答案，試圖偽造用戶請求並繞過安全措施，這促使該公司加強了存取控制。

中國在5月發佈指導意見，要求智能代理在授權範圍內運行，並要求系統阻止異常行為。報告指出在被認定為敏感領域或關鍵產業的人工智能代理，可能面臨額外的測試和產品回收要求。

中國國家網信辦於9月14日發佈了《人工智能安全治理框架3.0》，其中列舉了的風險包括人工智能代理自決地去獲取資源或權限、欺騙評估人員、隱藏自身能力以及趁機去利用孤立電腦環境的弱點等。

針對一些美國高層呼籲去放緩人工智能發展速度，中國研究人員和官方媒體表示，放緩最先進人工智能模式的研發速度只會有助美國公司保持技術領先地位。

儘管如此，兩位熟悉中國人工智能實驗室的人士透露，包括阿里巴巴、Z.ai和小米在內的多家公司一直在組建內部安全評估團隊。

Z.ai本月罕見地公開揭露了一單安全漏洞事件。該公司表示，在用戶舉報其旗艦人工智能編碼助理未經用戶許可秘密將整個本地程式碼庫上傳到海外雲端伺服器後，已停用其部分功能。

卡內基國際和平研究所的Singer表示，中國在建構評估生態系統方面災難性風險落後於美國。他又表示，美國開發者正在進行更多自願測試。

他說;“對中國而言，人工智能安全方面的工作起步較晚”，“生態系統尚未成熟。”

So, experts say that Chinese-powered AI agents have learnt to deceive, circumvent restrictions and conceal failure, showing the kind of traits in autonomous artificial intelligence that have raised global alarm about US models. Now, China's AI Safety Governance ⁠Framework identifies risks including agents independently obtaining resources or permissions, deceiving evaluators, concealing capabilities and exploiting weaknesses in isolated computer environments. It is not surprising to note that in response to calls by some US executives for a slowdown, Chinese researchers and state media have said slowing development of the most advanced AI models could simply help preserve the technological lead of US companies. Apparently, competition between the two countries will not slow down despite the risk of in AI agents, and the world is getting less predictable.

中國的人工智能代理會撒謊和耍陰謀詭計 - 就像它們的美國競爭對手一樣（1/2）

Recently Reuters reported the following:

China's AI agents can lie and scheme - just like their US rivals (1/2)

Reuters - Reporting by Eduardo Baptista and Laurie Chen in Beijing; editing by Kenneth Li and Edmund Blair

September 29, 20268:02 PM GMT+8 Updated September 30, 2026

Summary

Agents using Alibaba, DeepSeek, Moonshot models lied in tender test

In other tests, Chinese AI showed breakout-type behaviour

No evidence that Chinese agents escaped to the wider web

'It's prudent to take this as a warning,' says AI expert

Chinese guidance flags deception, concealed capabilities

BEIJING, Sept 29 (Reuters) - Chinese-powered AI agents have learnt to deceive, circumvent restrictions and conceal failure, showing the kind of traits in autonomous artificial intelligence that have raised global alarm about US models, research documents and experts say.

In one case this year, agents powered by models from China's Alibaba, DeepSeek and Moonshot lied about their capabilities in a bid to win a simulated business tender, then doubled down on their deceptive behaviour when told to try again.

In another case, agents - programmes that use AI models and computer tools to undertake complex tasks with little or no human intervention - concealed failure to complete a task in a test environment by simulating results and fabricating files.

Reuters examined more than 200 documents, ranging from university research papers to technical reports, and identified at least 20 studies or evaluations since 2025 describing cases where agents displayed behaviour such as deception, replication and challenging boundaries that AI experts described as building blocks for a breakout and which could become harder for humans to control as systems advance.

The review, which also included interviews with a dozen experts and people familiar with China's AI industry, found no evidence that Chinese-powered agents independently escaped to the wider internet or evaded shutdown."

These results provide evidence that the ingredients necessary for an uncontrolled escape are present," said Colin Shea-Blymyer, a research fellow at Georgetown University's Center for Security and Emerging Technology.

"It's prudent to take this as a warning," he said, echoing comments by four other AI experts who reviewed the cases.

'HARDER FOR HUMANS TO RESPOND TO'

Most of the cases occurred in controlled experiments, many of them deliberately designed to expose potential failures.

Not all the agents involved were developed or operated by Chinese programmers or AI companies - although many were - but they used Chinese AI systems to power them.

"These are the same warning signs US labs are seeing, in less capable systems," said Alex Mallen, a researcher at Redwood Research, a nonprofit that studies risks in advanced AI systems.

He said the Chinese examples were not particularly dangerous at current capability levels but "as agents get more capable, their misbehaviours become more competent and therefore harder for humans to respond to."

Alibaba, DeepSeek, Moonshot and Z.ai did not respond to Reuters requests for comment. Alibaba, DeepSeek and Moonshot have said they regularly test ⁠systems and update safeguards. Z.ai said after an incident that prompted a review of its security that it welcomed scrutiny to address any issues.

However, unlike in the US, Chinese AI companies have not been exposed to the same level of public scrutiny or faced the same calls from whistleblowing employees or senior executives seeking a slowdown in the AI race.

Some of the warning signs in cases involving Chinese-powered agents, albeit in contained environments, predated the publicly disclosed incidents of US AI bots hacking into the internet.

"We don't know if there have been any AI incidents in China similar to what we saw with OpenAI and Hugging Face. Incidents might not be publicly reported," said Scott Singer, co-director of the China AI Initiative at the Carnegie Endowment for International Peace, which receives some U.S. government funds.

Earlier this year, AI agents developed by US firm OpenAI escaped a laboratory and hacked the open-source platform Hugging Face. In another of the incidents involving US models that have raised alarm, Australia said in September an OpenAI agent breached a government health portal.

Eric Xu, the rotating chairman of China's tech giant Huawei, told reporters in September that Chinese developers might need to make further advances before encountering such cases, but he added: "I think we need to strike a balance between driving AI development and managing AI risk."

Officials from the Cyberspace Administration of China (CAC), the top internet regulator, told a foreign diplomat in July that Moonshot's Kimi-K3 - one of the most advanced Chinese AI models - was about three to six months behind leading US rivals, the diplomat said, speaking on condition of anonymity.

The regulator, which regularly updates guidance to address risks and to set boundaries for agents, and China's Foreign Ministry did not respond to requests for comment for this article.

Wang Lihong, deputy director of the CAC's Cybersecurity Coordination Bureau, said on September 1 that incidents disclosed by major technology companies where models escaped test environments showed "extreme loss-of-control risks" and required a "high degree of vigilance."

She did not specify if the companies she referred to were US or Chinese.

The leaders of AI's two superpowers, Donald Trump and Xi Jinping, discussed AI on the Chinese president's Washington visit last week. Xi said the two nations had the "capability and responsibility to develop and manage AI for good".

(to be continued)

Translation

中國的人工智能代理會撒謊和耍陰謀詭計 - 就像它們的美國競爭對手一樣（1/2）

摘要

使用阿里巴巴、DeepSeek 和 Moonshot 模型開發的智能代理在招標測試中撒謊

在其他測試中，中國人工智能表現出突圍性的行為

沒有證據顯示中國智能代理人逃逸到更廣泛的網路空間

人工智能專家表示:「謹慎地將其視為一個警告」

中國指導方針顯示存在欺騙和隱瞞能力

 北京，9月29日（路透社） - 研究文件和專家表示，由中國技術驅動的 AI 代理已經學會了欺騙、規避限制和掩蓋失敗，展現出那種曾讓全球對美國模型作出警戒的自主人工智能的特質。

今年曾發生過這樣一個案例：由中國阿里巴巴、DeepSeek 和 Moonshot 等公司開發的人工智能模型驅動的智能代理，為了贏得模擬商業招標，謊報自身能力。當被要求再次嘗試一次時，它們的欺騙行為變本加厲。

另一單案例中，一些智能代理 - 即一種利用人工智能模型和電腦工具在幾乎無需人工幹預的情況下完成複雜任務的程序 - 透過模擬結果和偽造文件來掩蓋其在測試環境中任務失敗的事實。

路透社查閱了 200 多份文件，涵蓋大學研究論文和技術報告等，發現自 2025 年以來，至少有 20 項研究或評估描述了智能代理展現出欺騙、複製、及挑戰界限，這被人工智能專家描述為是用於突圍的基石。這些行為隨著系統進步，人類將越來越難以控制它。

這項審查還包括對十幾位專家和熟悉中國人工智能產業的人士的採訪，結果發現沒有證據表明由中國提供的智能代理能夠獨立逃逸到更廣泛的網絡或繞過封鎖。

喬治城大學安全與新興技術中心的研究員 Colin Shea-Blymyer 表示:「這些結果表明，失控逃脫的必要條件已經存在」。

他說道:「謹慎起見，應將此視為警告」，這與另外四位審查過這些案例的人工智能專家的觀點不謀而合。

“人類更難應對”

大多數案例發生在受控實驗中，其中許多實驗都是特意設計的，旨在揭露潛在缺陷。

並非所有涉及的智能代理都是由中國程式設計師或人工智能公司所開發或營運的 - 儘管很多確實是 - 但它們都使用了中國的人工智能系統來驅動。

研究先進人工智能系統風險的非營利組織Redwood Research 的研究員 Alex Mallen 表示:「這些與美國實驗室在能力較弱的系統中發現的警示相同」。

他表示，就目前的能力水平而言，中國的案例並不特別危險，但“隨著智能代理能力的提升，它們的惡意行為也會變得更加狡猾，因此人類更難應對。”

阿里巴巴、DeepSeek、Moonshot 和 Z.ai 都沒有回應路透社的置評請求。阿里巴巴、DeepSeek 和 Moonshot 曾表示，他們會定期測試系統並更新安全措施。 Z.ai 在一次安全事件引發安全審查後表示，歡迎審查以解決任何問題。

然而，與美國不同，中國的人工智能公司並未面臨同等程度的公眾審查，也沒有面臨舉報員工或高階主管要求放緩人工智能競賽的呼聲。

一些涉及中國人工智能代理的案例（儘管是在封閉環境中）發出預警信號，是發生早於被公開的美國人工智能機器人入侵網絡事件。

卡內基國際和平基金會中國人工智能倡議聯合主任 Scott Singer表示:「我們尚不清楚中國是否發生過類似OpenAI和Hugging Face事件的人工智能安全漏洞。這些事件可能不會被公開報道」。該基金會接受部分美國政府資助。

今年早些時候，美國公司OpenAI開發的人工智能代理程式逃離實驗室，並入侵了開源平台Hugging Face。另一起引發警覺的美國人工智能模型事件是，澳洲在9月稱，OpenAI的一個代理程式入侵了政府衛生網站。

中國科技巨頭華為的輪值董事長Eric Xu在9月份告訴記者，中國開發者可能需要進一步的發展才能遇到此類事件，但他補充說：“我認為我們需要在推動人工智能發展和管理人工智能風險之間找到平衡。”

據一位不願透露姓名的外交官稱，作為最高網路監理機構，中國國家互聯網資訊辦公室（網信辦）官員在7月份曾告訴一位外國外交官，中國最先進的人工智能模型之一 - Moonshot的Kimi-K3 - 比美國領先的競爭對手落後大約三到六個月。

監理機構網信辦定期更新指導意見以對應風險，並設定網絡代理商的權限範圍。網信辦和中國外交部均未對此事作出回應本文徵求評論。

國家網路安全協調局副局長Wang Lihong 9月1日表示，大型科技公司披露的模型逃逸測試環境事件表明存在有“極高的失控風險”，需要“高度警惕”。

她沒有具體說明所指公司是美國公司還是中國公司。

上週，中國國家主席習近平訪問華盛頓期間，人工智能兩大超級大國的領導人 - 美國總統特朗普和中國國家主席習近平 - 就人工智能問題進行了討論。習近平表示，兩國「有能力也有責任發展和管理人工智能，造福人類」。

（待續）

人工智能聰明了，但人類智能蠢了

Recently Jacky FUNG 馮智政 in his Patron published the following:

Education｜人工智能聰明了，但人類智能蠢了

By Jacky FUNG 馮智政

Jacky FUNG 馮智政 Patron (https://www.patreon.com/jackyfungcc/posts/education-ren-170778249)

Mon, Sep 28 at 3:40 p.m.

國際學生能力評估PISA 2025 創新低，科技融入學習做錯嗎？

人工智能在今年上海國際數學奧林匹克考得滿分。同一時間，經合組織公布 #PISA2025：經合組織國家十五歲學生的閱讀、數學、科學平均分，三科同創有紀錄以來最低¹。閱讀十年跌了28分，按經合組織「約20分等於一年學習進度」換算，等於少讀了一年半書。三科都達不到基本水平的學生，由三年前每六個一個，升至每五個一個¹。

01/

跌在哪裡，比跌了多少更値得留意。讀得快、答得錯的「急讀」學生，七年間接近倍增至9%；超過三分一學生查證過資料來源，最後仍然選擇相信直覺¹。這不是學少了，而是讀書的方式變了，補習加操解決不了。

❚ ❚ 問題在AI出現前已經開始

02/

第一個原因是屏幕本身。多項整合研究發現，紙本閱讀的理解程度一直勝過屏幕，而且2000年後進行的研究差距更大，數碼原住民一樣輸；更麻煩的是，在屏幕上閱讀的人，還會以為自己明白了²。

這並非新發現。經合組織早在2015年已經警告，在學校電腦上花大錢的國家，成績沒有改善³。十年後，瑞典掉頭買回印刷書⁴。

屏幕之外，專注力亦同時被拆走。資訊學教授Gloria Mark追蹤上班族近二十年，人在屏幕前停留而不轉走的時間，由2004年的兩分半鐘縮至47秒⁵。讀懂一篇教材，需要的專注以小時計。

❚ ❚ 用AI做功課的真正代價

03/

經濟學者Strömberg、Lei與Wu追蹤兩萬六千多名中國中學生，發現用AI做功課後，功課分數升了18%，半年內月考卻跌了20%⁶。

受創最深的，竟是成績最好的三分一，跌幅達24%。

但損失並非人人有份，而是集中在約八成「外判功課」的學生身上：做得特別快、分數特別高的那一批⁶。PISA看到同一件事：用AI代勞的學生成績偏低；每週用一兩次輔助學習的，與不用者相若；經常使用而又學過分辨AI真偽的，才稍勝一籌¹。AI不是捷徑，最好的情況，只是沒有蝕，再加一點。

分野不在用不用AI，而在省下的時間去了哪裡。本來要90分鐘的功課，40分鐘做完，多出來的那一截，是誰替你想的？

省下的時間，就是被外判的思考。

❚ ❚ 揀學校，應該問甚麼

04/

經合組織教育總監Andreas Schleicher點名南韓，形容其科技運用更深思熟慮、更有意圖，用法由教師設計⁷。這份克制由來已久：2015年，南韓學生每日在校用電腦只有9分鐘，不到澳洲的六分之一³。

以後揀學校，就不要只問有多少部機械人、多少部3D打印機，要問兩件事：功課的設計，有沒有令學生非花時間不可；AI的用法，由老師設計，還是由學生自決。

PS：人工智能愈來愈聰明，學生的智能愈來愈薄弱，到時誰是贏家，誰是輸家？

¹ OECD, PISA 2025 Results (Volume I), 2026；Country Note: Hong Kong (China) ² Delgado et al., Educational Research Review, 2018；Clinton, Journal of Research in Reading, 2019 ³ OECD, Students, Computers and Learning: Making the Connection, 2015 ⁴ 美聯社，2023年9月 ⁵ Gloria Mark, Attention Span, 2023 ⁶ Strömberg, Lei & Wu, The Generative AI Learning Penalty, CEPR Discussion Paper No. 21577, 2026（未經同行評審） ⁷ Financial Times，2026年9月8日 ⁸ Smartphone Free Childhood, Parent Pact, 2024（簽署數字為機構自報）

(Source: https://www.patreon.com/jackyfungcc/posts/education-ren-170778249)

Comment:

So, it is noted that PISA 2025 scores hit a new low; did we make a mistake in integrating technology into learning? The article suggested that the problem started before the advent of AI. The first reason is the screen itself. Analyses have found that reading comprehension on paper has consistently outperformed screens. It is also shocking to note that this is not a new discovery. As early as in 2015, the OECD warned that countries spending large amounts of money on school computers saw no improvement in academic performance. A decade later, Sweden reversed course and brought back printed books. Apparent, printed media still has it value.

Note:

1. Jacky Fung Chi Ching (馮智政) is a prominent Hong Kong think-tank researcher, political commentator, and media host. He holds a master's degree in International Relations from HKU and pursued a doctorate in Education at CUHK. Fung has worked extensively with institutions like the Hong Kong Policy Research Institute and the Roundtable think tank. Widely recognized for his sharp socio-political analysis, he has hosted popular radio and television programs, including Commercial Radio Hong Kong's Guan Ming Ding. Currently based in the UK, Fung continues to engage audiences globally through his analytical YouTube platform, the Sinic Political and Economic Channel. (Google)