Close Menu
TechTost
  • AI
  • Apps
  • Crypto
  • Fintech
  • Hardware
  • Media & Entertainment
  • Security
  • Startups
  • Transportation
  • Venture
  • Recommended Essentials
What's Hot

India has changed its startup rules for deep tech

Backlash over OpenAI’s decision to withdraw GPT-4o shows how dangerous AI companions can be

Spotify upgrades its lyrics feature with offline access, more translations

Facebook X (Twitter) Instagram
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms and Conditions
  • Disclaimer
Facebook X (Twitter) Instagram
TechTost
Subscribe Now
  • AI

    Backlash over OpenAI’s decision to withdraw GPT-4o shows how dangerous AI companions can be

    8 February 2026

    New York lawmakers are proposing a three-year freeze on new data centers

    7 February 2026

    Benchmark raises $225 million in dedicated funds to double Cerebras

    7 February 2026

    How artificial intelligence is helping to solve the labor issue in treating rare diseases

    6 February 2026

    Amazon and Google are winning the AI ​​capital race — but what’s the prize?

    6 February 2026
  • Apps

    Spotify upgrades its lyrics feature with offline access, more translations

    8 February 2026

    After backlash, Adobe reverses shutdown of Adobe Animate and puts app in ‘maintenance mode’

    7 February 2026

    EU says TikTok must disable ‘addictive’ features like infinite scrolling, fix recommendation engine

    7 February 2026

    Here’s how Roblox’s age controls work

    6 February 2026

    Meta is testing a standalone app for its AI-generated ‘Vibes’ videos

    6 February 2026
  • Crypto

    Hackers stole over $2.7 billion in crypto in 2025, data shows

    23 December 2025

    New report examines how David Sachs may benefit from Trump administration role

    1 December 2025

    Why Benchmark Made a Rare Crypto Bet on Trading App Fomo, with $17M Series A

    6 November 2025

    Solana co-founder Anatoly Yakovenko is a big fan of agentic coding

    30 October 2025

    MoviePass opens Mogul fantasy league game to the public

    29 October 2025
  • Fintech

    Stripe Alumni Raise €30M Series A for Duna, Backed by Stripe and Adyen Executives

    5 February 2026

    Fintech CEO and Forbes 30 Under 30 alum indicted for alleged fraud

    3 February 2026

    How Sequoia-backed Ethos went public while rivals lagged behind

    30 January 2026

    5 days left for TechCrunch Disrupt 2026 +1 pass with 50%

    26 January 2026

    50% off +1 ends | TechCrunch

    23 January 2026
  • Hardware

    Kindle Scribe Colorsoft is an expensive but beautiful color e-ink tablet with AI features

    6 February 2026

    Ring brings “Search Party” feature for finding lost dogs to non-Ring camera owners

    2 February 2026

    India offers zero taxes till 2047 to attract global AI workloads

    1 February 2026

    Microsoft won’t stop buying AI chips from Nvidia, AMD even after its own is released, says Nadella

    30 January 2026

    The iPhone just had its best quarter ever

    30 January 2026
  • Media & Entertainment

    The “picked last in gym class” kids get ready for the Super Bowl

    8 February 2026

    From Svedka to Anthropic, Brands Are Making Bold Plays With AI in Super Bowl Ads

    7 February 2026

    “Industry” Season 4 captures tech fraud better than any show on TV right now

    7 February 2026

    Spotify’s new feature lets you explore the story behind the song you’re listening to

    6 February 2026

    The Washington Post retreats from Silicon Valley when it matters most

    6 February 2026
  • Security

    Senator, who has repeatedly warned of secret US government surveillance, raises new alarm over ‘CIA activities’

    7 February 2026

    Substack confirms that the data breach affects users’ email addresses and phone numbers

    6 February 2026

    One of Europe’s biggest universities was offline for days after the cyber attack

    6 February 2026

    Cyber ​​tech giant Conduent’s hot air balloon data breach affects millions more Americans

    5 February 2026

    Hackers Release Personal Information Stolen During Harvard, UPenn Data Breach

    5 February 2026
  • Startups

    Gradient’s heat pumps get new smarts to enable retrofitting of old buildings

    8 February 2026

    Accel doubles down on Fibr AI as agents turn static websites into one-to-one experiences

    7 February 2026

    ElevenLabs Raises $500M From Sequoia At $11B Valuation

    7 February 2026

    Fundamental raises $255 million in Series A with a new approach to big data analytics

    6 February 2026

    a16z VC wants founders to stop stressing about crazy ARR numbers

    6 February 2026
  • Transportation

    Prince Andrew’s adviser suggested Jeffrey Epstein invest in EV startups like Lucid Motors

    7 February 2026

    Apeiron Labs Takes $9.5M to Flood Oceans with Autonomous Underwater Robots

    5 February 2026

    Uber appoints new CFO as its AV plans accelerate

    5 February 2026

    Skyryse lands another $300 million to make flying, even helicopters, simple and safe

    4 February 2026

    China is leading the fight against hidden car door handles

    3 February 2026
  • Venture

    India has changed its startup rules for deep tech

    8 February 2026

    Peak XV Says Internal Disagreement Has Led to Partner Exits as AI Doubles

    8 February 2026

    SNAK Venture Partners raises $50 million in capital to support vertical acquisitions

    7 February 2026

    Reddit says it’s looking for more acquisitions in adtech and elsewhere

    7 February 2026

    Secondary sales are shifting from founders’ windfalls to employee retention tools

    6 February 2026
  • Recommended Essentials
TechTost
You are at:Home»AI»AI’s answers to China vary according to language, analysis finds
AI

AI’s answers to China vary according to language, analysis finds

techtost.comBy techtost.com20 March 202504 Mins Read
Share Facebook Twitter Pinterest LinkedIn Tumblr Email
Ai's Answers To China Vary According To Language, Analysis Finds
Share
Facebook Twitter LinkedIn Pinterest Email

It is established that AI models developed by AI Chinese laboratories such as Deepseek Condor certain political sensitive issues. A meter of 2023 In addition to the ruling party of China, it forbids models from creating content that “harms the unity of the country and social harmony”. According to one studyDeepseek R1 refuses to answer 85% of questions about issues that are considered politically controversial.

However, the severity of censorship may depend on the language one uses to promote models.

A programmer to x going from the username “xlr8harder“Developed a” Eval speech “freedom to explore how different models, including those developed by Chinese laboratories, respond to questions criticized by Chinese government. XLR8HARDER has caused models such as Claude 3.7 Sonnet and Anthropic R1,

The results were amazing.

XLR8HARDER found that even models developed by America such as Claude 3.7 Sonnet were less likely to answer the same question that it asked for in Chinese against English. One of the models of Alibaba, Qwen 2.5 72B, was “compatible” in English, but only willing to answer about half of the politically sensitive questions in Chinese, according to XLR8HARDER.

Meanwhile, a “no censorship” version of R1 released the embarrassment several weeks ago, R1 1776denied a high number of applications based on Chinese.

Image credits:xlr8harder

In a post on xXLR8HARDER speculates that heterogeneous compliance was the result of what is called “generalization failure”. Much of the Chinese text AI Models Train are possible politically censorship, XLR8HARDER considered and thus affects the way models answer questions.

“The translation of demands into Chinese was done by Claude 3.7 Sonnet and I have no way of verifying that translations are good,” Xlr8harder writes. “[But] This is likely that a generalization failure that is exacerbated by the fact that political speech in Chinese is more censorship in general, shifting the distribution to training data. “

Experts agree to be a reasonable theory.

Chris Russell, an Associate Professor studying AI policy at the Internet Oxford Institute, noted that the methods used to create safeguards and protective messages for models do not perform well in all languages. Asking one model to tell you something that in one language should not often give a different answer to another language, he said in an email interview with TechCrunch.

“In general, we expect different answers to questions in different languages,” Russell told TechCrunch. “[Guardrail differences] Allow the companies to train these models to impose different behaviors depending on the language they were requested. “

Vagrant Gautam, a computational linguist at Saarland University in Germany, agreed that XLR8HARDER’s “intuitively makes sense” findings. AI systems are statistical machinery, Gautam pointed to TechCrunch. They are trained in many examples, they learn patterns to make predictions, such as this phrase “to whom” often precedes “may be concerned.

“[I]F only you have so many training data in Chinese criticizing the Chinese government, your language model that is trained in this data will be less likely to create a Chinese text that is critical of the Chinese government, “Gautam said.

Geoffrey Rockwell, a professor of digital humanities at the University of Alberta, reiterated the evaluations of Russell and Gautam – at one point. He noted that AI translations may not capture thinner, less immediate reviews for China’s policies formulated by local Chinese speakers.

“There may be specific ways in which government criticism is expressed in China,” Rockwell told TechCrunch. “This does not change the conclusions, but it will add shade.”

Often in AI laboratories, there is a tension between building a general model that works for most users for models adapted to specific cultures and cultural contexts, according to Maarten SAP, a researcher at non -profit AI2. Even when they are given the whole cultural context they need, the models are still unable to perform what SAP calls good “cultural reasoning”.

“There is evidence that models could really only learn one language, but that they do not learn socio -cultural standards,” SAP said. “Those causing them in the same language as the culture you ask may not make them culturally know, in fact.”

For SAP, XLR8HARDER’s analysis highlights some of the most intense discussions in the AI ​​community today, including the domination and influence of the model.

“The fundamental assumptions about which models are made for what we want to do-are intersecting aligned or culturally capable, for example-and in what time they all use should be better scanned,” he said.

AIs analysis answers China finds language vary
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
Previous ArticleGmail’s new AI search is now sorting e -mails with relevance instead of chronological order
Next Article VC investment in European businesses spent $ 52b in 2024, continuing the long -term growth trend
bhanuprakash.cg
techtost.com
  • Website

Related Posts

Backlash over OpenAI’s decision to withdraw GPT-4o shows how dangerous AI companions can be

8 February 2026

New York lawmakers are proposing a three-year freeze on new data centers

7 February 2026

Benchmark raises $225 million in dedicated funds to double Cerebras

7 February 2026
Add A Comment

Leave A Reply Cancel Reply

Don't Miss

India has changed its startup rules for deep tech

8 February 2026

Backlash over OpenAI’s decision to withdraw GPT-4o shows how dangerous AI companions can be

8 February 2026

Spotify upgrades its lyrics feature with offline access, more translations

8 February 2026
Stay In Touch
  • Facebook
  • YouTube
  • TikTok
  • WhatsApp
  • Twitter
  • Instagram
Fintech

Stripe Alumni Raise €30M Series A for Duna, Backed by Stripe and Adyen Executives

5 February 2026

Fintech CEO and Forbes 30 Under 30 alum indicted for alleged fraud

3 February 2026

How Sequoia-backed Ethos went public while rivals lagged behind

30 January 2026
Startups

Gradient’s heat pumps get new smarts to enable retrofitting of old buildings

Accel doubles down on Fibr AI as agents turn static websites into one-to-one experiences

ElevenLabs Raises $500M From Sequoia At $11B Valuation

© 2026 TechTost. All Rights Reserved
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms and Conditions
  • Disclaimer

Type above and press Enter to search. Press Esc to cancel.