OpenAI unveils ChatGPT Health, says 230 million users ask about health each week

Key Takeaway: OpenAI introduced ChatGPT Health, a separate space for health conversations that can connect to wellness apps and medical records—while claiming it won’t use those chats to train its models.

More Insights:

• OpenAI says 230 million users ask health and wellness questions on ChatGPT each week.

• Health chats will be siloed from your normal chats, so sensitive context won’t bleed into everyday conversations.

• If you discuss health elsewhere, ChatGPT may nudge you into the Health section.

• Health can still draw on some context from the standard experience (e.g., knowing you’re training for a marathon).

• OpenAI also acquired Torch (reportedly for $100M in equity) to help unify medical data into a “medical memory for AI.”

Why it matters: This is a bet that the biggest healthcare bottleneck isn’t information—it’s continuity and access—and whoever becomes the trusted “health context layer” could reshape how people make everyday medical decisions, for better or dangerously worse if accuracy and accountability don’t keep up.

Related: Anthropic announces Claude for Healthcare following OpenAI’s ChatGPT Health reveal

Join the Most In-Depth LIVE AI Training & Master AI in 2 Days (Free for 48h)

It's January again, and you're probably making New Year’s resolutions. Because you want 2026 to be better, but the smartest people aren't just planning. They're already mastering AI, the skill that matters most this year. Around 48% of workers globally used AI tools in 2025, and 2026 is when AI truly peaks.

Outskill can teach you AI in just 2 days - the skill that can make you smarter, faster, more revenue-driven, and absolutely irreplaceable in 2026.

• 🧠 Live sessions- Saturday and Sunday

• 🕜 10 AM EST to 7PM EST

• 🎁 You will also unlock $5000+ in AI resources: (as bonuses when you show up!)

Apple picks Google’s Gemini AI for its big Siri upgrade

Key Takeaway: Apple is partnering with Google to use Gemini models and cloud tech to power future AI features like Siri, aiming to close the gap with rivals while keeping Apple’s privacy-first approach.

More Insights:

• Apple says Google’s tech is the “most capable foundation” for Apple’s next-gen foundation models.

• The deal is multi-year and non-exclusive, leaving room for other AI partners.

• Reports suggest Apple may pay around $1B, after also testing OpenAI and Anthropic.

• Apple Intelligence started rolling out in 2024, focusing on subtle, OS-level AI and on-device/privacy-controlled processing.

• A more personalized Siri has been delayed, but Apple says an upgrade is coming this year, with reports pointing to a spring launch.

Why it matters: Apple choosing an external AI engine signals a strategic shift: even the most vertically integrated tech company is acknowledging that “best AI” may require partnering—yet doing so under heavy antitrust pressure and privacy expectations could determine whether AI becomes a trusted default layer of computing or just another feature users resent.

Anthropic cracks down on unauthorized Claude usage by third-party harnesses and rivals

Key Takeaway: Anthropic rolled out new technical and contractual enforcement to stop third-party tools from spoofing Claude Code (to dodge API pricing/limits) and to restrict competitors like xAI from using Claude via Cursor for competitive model development.

More Insights:

• Anthropic tightened safeguards against “Claude Code harness” spoofing, disrupting tools like OpenCode and triggering some mistaken auto-bans (now being reversed).

• The core issue: third-party wrappers used consumer subscriptions (Pro/Max) to run high-intensity autonomous coding loops that resemble API-scale usage.

• Developers framed it as a “buffet vs. speed limit” problem: the subscription is all-you-can-eat, but Claude Code controls pacing—harnesses remove the brakes.

• OpenCode responded with a new paid tier that routes through an enterprise API gateway, and signaled plans to integrate OpenAI/Codex subscriptions.

• Separately, Anthropic restricted competitor access (xAI) tied to commercial terms prohibiting using Claude to build/train competing products, with Cursor being the access path.

Why it matters: This is the moment “agentic dev” grows up—AI labs are drawing hard boundaries between human subscription use and automation at scale, and teams betting on clever workarounds are discovering a brutal truth: if your workflow depends on unofficial access paths, your “cost savings” can disappear overnight, taking production velocity with it.

Other stuff

• OpenAI Sweetpea Earbuds Leak

• Google announces a new protocol to facilitate commerce using AI agents

• ‘I rarely get outside’: scientists ditch fieldwork in the age of AI

• Mark Zuckerberg says Meta is launching its own AI infrastructure initiative

• Hands-on with Bee, Amazon’s latest AI wearable

• Musk’s X could be banned in Britain over AI chatbot row

• One Game Publisher Is Pushing Back Against AI Art With Contracts That Ban It: ‘It’s Cancerous’

Hope you enjoyed today's newsletter

⚡️ Join over 300,000 people using the Superpower ChatGPT extension on Chrome and Firefox.