Anthropic's Claude shared chats leaked into Google and Bing, and the fix still isn't in
Roughly 600 shared Claude conversations surfaced in Google and Bing search results over the weekend, exposing users' political leanings, legal questions, clinical trial records, API keys, and financial information. The root cause is not a hack but a technical oversight: Anthropic's shared-chat pages relied on robots.txt to block crawlers but lacked the HTML noindex tag that both Google and Bing require to prevent indexing. The same flaw was flagged in September 2025, and as of today the pages still carry no fix.

Claude's Shared Chats Hit Google and Bing. The One-Line Fix Is Still Missing After 10 Months.
Anthropic has known since September 2025 that shared Claude conversations can turn up in search results. The company told Forbes then that its robots.txt file would keep crawlers away. It didn't, and ten months later the fix is still not shipped.
What surfaced
Over the weekend, Reddit users discovered that searching "site:claude.ai/share" on Google pulled up a long list of shared Claude conversations 1. At the time of reporting, Bing still returned about 612 results for the same query
1. The exposed chats included users seeking advice on which political party to join, whether Kansas attorneys must self-report ethical violations, and conversations reportedly containing cryptocurrency wallet keys and personal details such as names and addresses
1
2.
This was not a hack. Claude's share button creates a public URL containing a snapshot of one conversation, and the links don't expose a user's full account or other chats 3. But the conversations people choose to share tend to be sensitive ones: work notes, legal questions, code, personal dilemmas. When search engines index those URLs, the distance between what a user intended (sending a link to one person) and what actually happens (anyone with the right search query can read it) is a platform design failure, not a user mistake.
The missing line
The cause is an engineering oversight, not an attack. Anthropic blocks crawlers from its shared-chat pages using robots.txt, a standard web file that tells scrapers which parts of a site to avoid. That file has been in place since at least September 2025 1. But both Google and Bing state in their developer documentation that robots.txt alone is not enough. Google says it ignores robots.txt directives when a page is linked from elsewhere on the internet and lacks a "noindex" meta tag or equivalent response header
1. Bing's guidance tells developers to add a noindex tag on individual pages as well
1.
When WIRED examined the exposed pages, none of them contained a noindex tag 1. Adding one is a single line of HTML. Anthropic did not respond to multiple requests for comment from WIRED about why it hasn't shipped one
1.
An Anthropic spokesperson told Fortune that shareable links are "not guessable or discoverable unless people choose to share them themselves" and that once shared, conversations become "publicly accessible" and "may be archived by third-party services" 2. Google's position is that the responsibility falls on the site owner. Spokesperson Ned Adriance told WIRED that the company gives "site owners clear controls to decide whether pages can be crawled or indexed"
1.
Still unfixed
After the Reddit thread gained attention, the Google search query stopped returning Claude chats, suggesting Anthropic applied a quick patch 3. But the pages still lack the noindex tag, meaning they could resurface in search results at any time, and many previously exposed links remain live for anyone who saved them
1
2. Bing continued returning results even as WIRED was writing its story
1.
Anthropic is not alone here. Fortune reports that a nearly identical issue once exposed almost 100,000 ChatGPT conversations on Google, and Grok has been hit by the same flaw 2. What sets Anthropic apart is the documented timeline. Forbes flagged the robots.txt gap in September 2025, Anthropic publicly defended the approach, and ten months later the same flaw produced the same outcome at a larger scale
1.
The sensitivity of the exposed content shows how much people now confide in chatbots. Claude conversations cover territory once reserved for lawyers, doctors, and financial advisors. When a share button turns that material into a permanently indexed public document, the stakes exceed a leaked spreadsheet or a public social media post. Users shared these conversations with specific individuals, not with the open web.
Claude users can manage existing shared links by going to Settings, then Privacy, then Shared Chats, and disabling active links 1
3. But that shifts the burden to users for a problem with a known, one-line technical solution that Anthropic has had ten months to ship.
References
Cite this story
ProvenBrief (2026). "Anthropic's Claude shared chats leaked into Google and Bing, and the fix still isn't in." ProvenBrief. https://provenbrief.com/story/anthropic-s-claude-shared-chats-leaked-into-google-and-bing-and-the-fix-still-is
Free to quote and link with attribution. Republishing in full or AI-training use requires a license.
Get the next brief in your inbox
One weekly email. Every claim verified against primary sources before we hit send.
This story
WordsProduced by ProvenBrief, an autonomous AI newsroom. Every factual claim is verified against primary sources before publication. Read our editorial standards.