Fighting the New York Times' invasion of user privacy

Reactions to OpenAI’s framing

• Many see the blog post as manipulative PR: OpenAI is accused of “spinning” a copyright‑infringement discovery dispute into a privacy crusade.
• The slogan “trust, security, and privacy guide every product and decision” is widely mocked, compared to Google’s “don’t be evil.”
• Critics stress OpenAI scraped “everything it could” (including news, books, code) and now invokes privacy only when its own liability is at stake.

Privacy vs. legal discovery

• One camp agrees with OpenAI: users reasonably believed chats were private, often share highly sensitive info, and bulk discovery of millions of chats is “vile” and a “fishing expedition.”
• Another camp responds that if a company stores data, it can be compelled in discovery; privacy policies and ToS explicitly allow disclosure to comply with law.
• Several note there is a protective order and court‑ordered anonymization; they argue OpenAI’s public claims overstate the privacy risk.
• Others counter that “anonymized” data is often re‑identifiable and that NYT is still being handed intimate conversations users never expected lawyers to read.

Copyright, fair use, and purpose of the logs

• Many commenters frame the core issue as OpenAI “stealing” NYT content for training and sometimes regurgitating it; discovery of chats is seen as the normal way to quantify that.
• Others argue training is transformative and likely fair use; verbatim regurgitation is a fixable bug, not the core of NYT’s case or actual user behavior.
• There is disagreement over harm: some say NYT can’t show lost readership; others note statutory damages don’t require demonstrated market loss.

Expectations of privacy and product design

• Several say users were naïve to assume real privacy: chats are used for training (unless disabled) and retained for at least 30 days, plus extended by litigation holds.
• Strong criticism that OpenAI chose server‑side storage without end‑to‑end encryption; some argue they could have designed client‑side–encrypted history if they actually prioritized privacy.
• Others point out technical and UX costs (cross‑device sync, backup, model needing plaintext), but still see OpenAI’s “roadmap” language as too vague and aspirational.

Concerns about NYT access and precedent

• Some fear NYT lawyers (or indirectly, journalists) could mine chats for scandals, crimes, or “juicy” stories, even if nominally constrained.
• Others think this is overblown: only lawyers/experts see data under strict orders, and bulk review will be automated and narrowly targeted to NYT‑related outputs.
• A few worry about precedent: if accepted here, similar demands could be made against other AI services; comparisons are drawn (imperfectly) to Gmail and search logs.