Claude

37 cities are each taking on one hard question about AI this month. These answers came from New York. We spent an evening in small groups, digging into the hopes, the worries, and the asks people have for each other and for the people building AI.

The information you share with AI only travels as far as you let it. Zoe from the Anthropic education team explains what happens to your information when you chat with an AI, how long it stays there, and how to take control of where it goes. Have a question? Let us know in the comments.

Claude Tag now follows the whole channel for proactive contributions. In Slack channels where you've added it, @Claude replies when @-mentioned and can join in when it judges it can help. Until now it made that call one message at a time. Now it weighs the whole conversation, its memory, and your standing instructions to make one of four moves: reply inline, start deeper work in a thread, route the message to work already in flight, or say nothing. The result is a Claude that's roughly 30% better at judging when to contribute and when to stay quiet, always within the permissions, tools, and scope you've configured. In most channels, on most messages, the right move is saying nothing. Admins choose which channels Claude can post in, and any channel member can switch 'Respond automatically' off so it replies there only when @-mentioned. This update comes at no additional cost today. While holding more context does increase Claude Tag’s usage, the additional context Claude Tag holds does not count toward usage or spend limits on any plan. Available in beta for Claude Enterprise and Team plans in Slack. Read more: https://lnkd.in/eDXuKaMC

We've rebuilt the Claude in Chrome side panel as a Claude Cowork session, with the same conversation history, skills, and connectors as Claude on desktop, web, and mobile. Sessions live with your account, not on any single device. Start a task in a browser tab and pick it up on your desktop or phone, with the whole conversation saved to your history. The new side panel is on the Max and Team plans today, and rolling out to the Pro plan in the coming weeks. Enterprise plan admins can enable it for their orgs. Browser agents can be tricked by instructions hidden in a page. We build defenses against this, and we still recommend a few habits of your own: https://lnkd.in/eVZqRyCv Give it a try: claude.com/chrome

How much you can trust an AI depends on what you’re asking. Kyra from the Anthropic education team breaks down the two most common reasons for an AI to be confidently wrong: hallucination and sycophancy. Have a question? Let us know in the comments.

Twenty-five years ago, Dr. Patrick Walker founded Arkana Labs with the belief that patients shouldn't have to wait for answers. Today, most renal pathology cases in the country come through a single lab in Little Rock that calls itself an ER for tissue. Arkana reads roughly 160 cases a day. Each one arrives with paperwork as long as 90 pages, which someone used to thumb through page by page. Now they use Claude to triage cases so physicians can spend their time on the work only they can do. Over those 25 years, the lab has accumulated decades of kidney tissue and diagnostic reports. Arkana is using Claude to turn that archive into structured data for training diagnostic models that could catch kidney disease before a biopsy is ever needed. "We'd love to make ourselves irrelevant as clinicians and more relevant as researchers, helping in drug discovery. That's the ultimate goal." —Owen Parker, CTO Read more: https://lnkd.in/emztmmZK

In late 2025, Iceland’s government launched one of the world’s first national AI education pilots, giving teachers who volunteered access to AI tools. We visited Iceland to find out how people are thinking about AI. More information about the pilot: https://lnkd.in/eKevAUtr