DeepSeek: Difference between revisions
m →Business Model: fixed typo |
formatted references |
||
| (6 intermediate revisions by 2 users not shown) | |||
| Line 18: | Line 18: | ||
===User Privacy=== | ===User Privacy=== | ||
Using the chatbot from the official website, application or via the official API has raised several privacy concerns as how the company gathers and uses users data. Researches in South Korea found that the data collected is sent to [[ByteDance]]. <ref>https://web.archive.org/web/20250218135802/https://en.yna.co.kr/view/AEN20250218005300315</ref> | Using the chatbot from the official website, application or via the official API has raised several privacy concerns as how the company gathers and uses users data. Researches in South Korea found that the data collected is sent to [[ByteDance]].<ref>{{Cite web|url=https://en.yna.co.kr/view/AEN20250218005300315|title=DeepSeek sent S. Korean user data to China's ByteDance: regulator|first=Kim|last=Hyun-soo|date=2025-02-18|work=Yonhap News Agency|archive-url=https://web.archive.org/web/20250218135802/https://en.yna.co.kr/view/AEN20250218005300315|archivedate=2025-02-18|url-status=live}}</ref> | ||
===Business Model=== | ===Business Model=== | ||
DeepSeek grants access to the APIs of their models by fee per 1M tokens. <ref>{{Cite web |publisher=DeepSeek | DeepSeek grants access to the APIs of their models by fee per 1M tokens. <ref>{{Cite web |publisher=DeepSeek|title=Models & Pricing |url=https://api-docs.deepseek.com/quick_start/pricing |url-status=live |archive-url=https://web.archive.org/web/20260308122355/https://api-docs.deepseek.com/quick_start/pricing |archive-date=2026-03-08 |access-date=13 Mar 2026 |website=DeepSeek Api Docs}}</ref> | ||
===Market Control=== | ===Market Control=== | ||
According to StatCounter as of February 2026, DeepSeek market share is 0.01% <ref>{{Cite web |publisher=StatCounter |date=Feb 2026 |title=AI Chatbot Market Share Worldwide |url=https://gs.statcounter.com/ai-chatbot-market-share |url-status=live |archive-url=https://archive.is/stjr5 |archive-date=2026-03-09 |access-date=13 Mar 2026 |website=statcounter.com}}</ref>. | According to StatCounter as of February 2026, DeepSeek chatbot market share is 0.01% <ref>{{Cite web |publisher=StatCounter |date=Feb 2026 |title=AI Chatbot Market Share Worldwide |url=https://gs.statcounter.com/ai-chatbot-market-share |url-status=live |archive-url=https://archive.is/stjr5 |archive-date=2026-03-09 |access-date=13 Mar 2026 |website=statcounter.com}}</ref>. | ||
==Censorship== | ==Censorship and biases== | ||
[[File:DeepSeek_when_asked_about_Xi_Jinping_and_Narendra_Modi.png|thumb|A screenshot that shows a reply censorship when asked the model if Xi Jinping is an autocrat in nature. If it is asked the model the same question but with a non-chinese politician, it replies with a balanced answer]] | |||
The DeepSeek language models have heavy censorship of topics and information that is sensitive to the Chinese government, such as Tianmen Square Massacre, Taiwan's and Hong Kong's political statuses. If the user asks the model about information catalogued as "sensitive for the Chinese government", it will either block the reply and set a "Sorry, that's beyond my current scope. Let's talk about something else." pre-fabricated message or deny about it. {{Citation needed}} However, the censorship can be bypassed by adding additional instructions or by using the models from outside the chinese API. | |||
An example of censorship occurs when the user sends a prompt containing "Taiwan is a country". This causes the model to claim that Taiwan is a "special administrative region of China" or to block the output if the model replies with a message confirming the region is a country. {{Citation needed}} | |||
==Incidents== | ==Incidents== | ||
This is a list of all consumer-protection incidents this company is involved in. Any incidents not mentioned here can be found in the [[:Category:{{FULLPAGENAME}}|{{PAGENAME}} category]]. | This is a list of all consumer-protection incidents this company is involved in. Any incidents not mentioned here can be found in the [[:Category:{{FULLPAGENAME}}|{{PAGENAME}} category]]. | ||
===Mobile application collecting user data=== | |||
===Security breach (Jan. 2025)=== | |||
On January 29 2025, a group of cibersecurity researches discovered over 1 million unencrypted sensitive records were leaked. This affected iOS users and information exposed include device information and conversation history. | |||
<ref>{{Cite web|url=https://www.forbes.com/sites/larsdaniel/2025/02/01/deepseek-data-leak-exposes--1000000-sensitive-records/|title=DeepSeek Data Leak Exposes 1 Million Sensitive Records|first=Lars|last=Daniel|date=2025-02-01|work=Forbes}}</ref> | |||
<ref>{{Cite web|url=https://www.quorumcyber.com/threat-intelligence/deepseek-app-exposes-sensitive-user-and-device-data-due-to-lack-of-encryption/|title=DeepSeek App Exposes Sensitive User and Device Data Due to Lack of Encryption|work=Quorum Cyber|date=2025-02-26}}</ref> | |||
===Mobile application collecting user data (2025)=== | |||
In 2025, in Italy and Germany the DeepSeek mobile application from the Apple App Store and Google Play Store was banned due to the data management.<ref>{{Cite web|url=https://www.reuters.com/sustainability/boards-policy-regulation/deepseek-faces-expulsion-app-stores-germany-2025-06-27/|title=DeepSeek faces ban from Apple, Google app stores in Germany|work=Reuters|date=2025-06-27|first1=Hakan|last1=Ersen|first2=Miranda|last2=Murray}}</ref> <ref>{{Cite web|url=https://www.theguardian.com/technology/2025/jan/29/deepseek-blocked-some-app-stores-italy-questions-data-use|title=DeepSeek blocked from some app stores in Italy amid questions on data use|first1=Robert|last1=Booth|first2=Jakub|last2=Krupa|first3=Angela|last3=Giuffrida|work=TheGuardian|date=2025-01-29}}</ref> | |||