What privacy red flags exist in DeepSeek terms?

TL;DR

DeepSeek raises several privacy concerns: data could be collected broadly including text, audio, and chat history with unclear retention and no explicit model training opt-out. Data may be stored in China and shared with advertisers or analytics partners, raising compliance and user-rights questions. Workarounds include offline local models or alternative services.

Transcript

if you use the Deep seek website I think it's important that you understand their privacy policy and their terms of use because there is at least three very alarming things that I found in there when I was going through those documents and I wanted to share this in this video because I'm making another video that I'm posting right after this one wh... Read More

Key Insights

• The policy lists extensive user data collection including input, chat history, and feedback.
• The document does not specify data retention timeframes or anonymization for analyzed data.
• There is mention of keystroke patterns as a form of biometric data.
• Data sharing with advertisers and analytics partners is stated with no explicit opt-out.
• Data storage is described as being in China, triggering cross-border compliance concerns.
• EU GDPR and other international privacy laws are not clearly addressed in the policy.
• There is no clear user-rights mechanism for full data deletion or data portability.
• The video suggests offline local deployment of DeepSeek as a privacy workaround.

Questions & Answers

Q: A real search query about the video: What privacy concerns does DeepSeek raise in its policy?

The video identifies several privacy concerns in DeepSeek policy and terms. It notes extensive data collection including text input, audio, uploaded files, chat history and feedback. It highlights lack of clarity on data retention, potential model training use, and no explicit opt-out for data sharing with advertisers or analytics partners. It also points out biometric data like keystroke patterns and data storage in China, with implications for GDPR and other laws.

Q: How does the speaker suggest reducing privacy risk with DeepSeek?

The speaker suggests using DeepSeek offline locally to avoid sending data to the cloud, describing methods to run distilled DeepSeek R1 models on a computer or using alternative interfaces like perplexity that host the model locally. This approach eliminates internet traffic and reduces exposure to data sharing, though it may limit features and require technical setup.

Q: What specific data collection concerns are highlighted as problematic?

Specific concerns include extensive user input collection, keystroke rhythm data, chat history, and uploaded files. The lack of explicit anonymization, retention timelines, and clarity on whether data is used to train models are cited as problematic. The policy also mentions device IDs and cross-device tracking, increasing the privacy risk.

Q: What is said about data storage location and its implications?

The video states that the policy indicates data is stored in servers located in China. This raises concerns about compliance with international privacy laws and government access under Chinese law. The lack of clear safeguards for EU users or other regions compounds worries about cross-border data transfers and protection standards.

Q: Is there any mention of user control over data or deletion rights?

The speaker notes there is no explicit option for users to prevent data use or a clear path to full data deletion. The policy mentions data retention generally but does not provide a concrete deletion timeframe or a straightforward process for users to request deletion or data portability.

Q: What updates to DeepSeek policies are mentioned and why matter?

The video references updates to the privacy policy updated December 05, 2024 and the terms of use updated January 20, 2025. These updates matter because they may change data collection practices, retention, sharing, and user rights, so ongoing review is necessary for users to understand current protections and obligations.

Q: What alternatives or workarounds are proposed for privacy-conscious users?

Alternatives include downloading and running DeepSeek R1 locally from sources like ama.com to operate offline, or using Perplexity Pro’s hosted reasoning with R1 to keep processing on local servers. Both approaches reduce or avoid sending data to the DeepSeek cloud, but may involve complexity and potential feature limitations.

Q: What general caution does the video offer regarding DeepSeek for sensitive use?

The video advises caution for using DeepSeek for sensitive, legal, or business-critical purposes due to questionable privacy protections and unclear compliance with global privacy laws. It recommends considering alternative providers with stronger privacy protections and clearer user-rights safeguards before relying on DeepSeek for such tasks.

Summary & Key Takeaways

• The video highlights that the policy covers extensive data collection including profile data, user input, and automatically collected data, without clear retention timelines or anonymization details, raising privacy concerns. It also notes potential data sharing with third parties and the possibility of data storage in China, which implicates international law and compliance.

• The presenter contrasts the DeepSeek privacy policy with alternatives and describes ways to use DeepSeek offline or via other services to mitigate data exposure, stressing caution for sensitive uses and preferences for stronger privacy protections.

• Additionally, the video references a December 05, 2024 update and a January 20, 2025 update to the terms, indicating ongoing changes and the need for continual policy review when using the service.