South Korean regulator accuses DeepSeek of sharing user data with ByteDance

Original Article ↗ Hacker News Discussion ↗

Scope of the Allegation

  • Some commenters stress the difference between:
    • An earlier incident where DeepSeek’s database was “unintentionally exposed” to the public internet.
    • The current accusation of deliberate sharing of user data with ByteDance via embedded SDKs/analytics.
  • Others argue both are serious privacy failures and illustrate the need for stronger data‑protection laws and individual control over data.

“Everyone Does It” vs. Special Risk from China

  • Many compare DeepSeek–ByteDance links to ubiquitous use of Google Analytics, Firebase, Crashlytics, Meta pixels, etc., where user/device data is routinely sent to U.S. tech giants.
  • One view: using ByteDance analytics is “a Chinese company using a Chinese cloud,” analogous to Western apps using Google/Meta; outrage is misplaced or hypocritical.
  • Opposing view: Chinese firms are different because of state access and national‑security concerns; critics emphasize interconnected Chinese companies, links to entities like China Mobile, and existing bans (TikTok, DeepSeek in South Korea).
  • There is debate over whether this is genuine security concern or protectionism/ethnonationalism, with some explicitly calling out US/China double standards.

What Data Is Actually Shared?

  • Technical commenters explain that the sharing appears to be:
    • Ongoing analytics/configuration data from the mobile app (device metadata, behavior, performance, crashes).
    • Possible, but unproven, inclusion of raw chat content.
  • A separate iOS analysis reports poor security practices (hardcoded keys, legacy crypto, some HTTP traffic), raising general trust issues but not a clear “smoking gun.”

Model Bias, Propaganda, and Censorship

  • Several discuss DeepSeek’s alleged built‑in censorship/propaganda:
    • Some models reportedly refuse topics like Tiananmen Square; others and distilled/offline variants are described as relatively uncensored or easy to jailbreak.
    • Counterpoint: Western AIs also embed ideological filters (“woke” complaints), so concerns about Chinese censorship are framed by some as a double standard.
  • For many users, practical impact is weighed: if the tool is for coding or mundane tasks, geopolitical censorship may feel irrelevant; for others, any state‑aligned narrative control is unacceptable.

Trust in Big Tech and Messaging Platforms

  • Side debate over whether WhatsApp/Meta really respect end‑to‑end encryption, with:
    • One side demanding concrete evidence before assuming backdoors.
    • The other arguing Meta’s history removes the “benefit of the doubt” and that skepticism is rational.