According to @OpenAI, ChatGPT Health keeps health chats, files, and memories in a separate dedicated space, with health conversations visible in history but their information not flowing into regular chats, source: OpenAI on X, Jan 12, 2026. Users can view or delete Health memories anytime in Health or via Settings > Personalization, source: OpenAI on X, Jan 12, 2026. No cryptocurrencies or blockchain integrations are mentioned in the announcement, source: OpenAI on X, Jan 12, 2026.

According to @paoloardoino, Tether announced its first AI launch featuring QVAC Workbench, a mobile and desktop app to run and experiment with multiple AI models locally with 100% privacy (source: https://twitter.com/paoloardoino/status/1981715700260450738). The announcement also introduced QVAC Genesis I, described as the largest synthetic pre-training datasets for LLMs to date (source: https://twitter.com/paoloardoino/status/1981715700260450738; source: https://x.com/Tether_to/status/1981693435208695917). The posts do not provide token launches, pricing, release timelines, or blockchain integrations (source: https://twitter.com/paoloardoino/status/1981715700260450738; source: https://x.com/Tether_to/status/1981693435208695917). For trading context, the update confirms Tether’s entry into on-device AI tooling and LLM data infrastructure under the QVAC brand while offering no direct market action items in the announcement (source: https://twitter.com/paoloardoino/status/1981715700260450738).

According to Jeff Dean, Google Research released VaultGemma, an open large language model trained from scratch with differential privacy, positioning it as a privacy-preserving foundation for AI applications (source: Jeff Dean on X; Google Research blog). The official blog describes VaultGemma as the world’s most capable differentially private LLM and links a technical report introducing a scaling law for differentially private language models, informing how DP-LM performance scales under privacy constraints (source: Google Research blog; arXiv:2501.18914). For crypto and DeFi trading teams integrating LLMs with sensitive datasets and keys, differentially private training is designed to limit memorization and exposure of individual training examples, reducing data leakage risks in AI-driven workflows (source: Google Research blog; arXiv:2501.18914). No token price or trading volume impacts were disclosed alongside the release; this is a technical update with direct relevance to privacy-preserving AI tools used in Web3 operations (source: Jeff Dean on X; Google Research blog).

According to Sam Altman (@sama), the need for 'AI privilege'—treating conversations with AI like those with lawyers or doctors—has become urgent and could accelerate regulatory shifts. This development signals potential changes in how user data, privacy, and compliance are handled in AI-driven cryptocurrency platforms, which may impact trading strategies and risk assessment for crypto investors (source: @sama on Twitter, June 6, 2025).