Anthropic's Claude Sharing Bug Let Google Index Chats, Exposing Crypto Seed Phrases

AI Market Summary
A Claude sharing misconfiguration allowed Google (and still some Bing results) to surface publicly indexed chats and artifacts, including exposed crypto seed phrases and other credentials. While Anthropic has added noindex controls and Google removals began, third-party archives persist, raising ongoing compromise risk. The incident can increase near-term security concerns, reduce user trust in cloud AI tooling for sensitive workflows, and elevate operational risk across crypto holders and teams.
Impact level
● Medium
Affected assets
BTC/USDT-2.30%
AI Insight · BTC/USDTAI Insight
▼ Bearish
Trade now
⚠️ AI-generated insights are based on news content and are provided for informational purposes only. They do not constitute investment advice or represent the views of BingX. Investing involves risk. Please trade responsibly.
Anthropic's Claude inadvertently made some users' shared chats searchable on Google, and a portion of those conversations contained highly sensitive crypto information. What happened Claude share links are labeled "Anyone with the link," implying unlisted, link-only access. A Reddit user discovered that Google was indexing those URLs. On July 25, a site-restricted query (site:claude.ai/share) surfaced hundreds of conversations; the accompanying Reddit post drew about 4,000 upvotes. The indexed results ranged from innocuous prompts to material that should never be public: a pasted crypto wallet seed phrase (effectively the master key to a wallet), legal questions tied to ethics concerns, internal spreadsheets, product roadmaps, and other explicit or unusual queries. One widely shared example read: "Why is the record industry hiding the fact that birds are secretly professional rappers…?" The underlying issue was a small but critical omission: Claude's share pages lacked an HTML noindex tag. Anthropic intended the feature to function like an unlisted video, but without noindex, search engines could index any share URL they discovered elsewhere online. Anthropic had a robots.txt rule blocking share URLs, but that alone did not prevent indexing. Google can still index a URL it finds on other sites even if its crawler can't fetch the page content. In those cases, Google may display "No information is available for this page" while still linking to the exposed URL. The exposure went beyond chats. Users also found publicly indexed "artifacts" (via queries such as site:claude.ai/public/artifacts), including payroll spreadsheets, CRM exports, and unreleased product roadmaps. By July 26, Google began removing Claude results after Anthropic added noindex tags and adjusted its configuration. Bing was still returning some shared links. Separately, a public GitHub repository titled SharedClaudeChats has already archived 453 Claude conversations and 519 Grok chats—11,241 messages in plain text—suggesting copies of exposed content may persist outside Anthropic's control. Anthropic had not issued a public statement at the time of publication. The episode follows other recent missteps, including requesting government IDs from users in April and quietly removing hidden tracking code in July after it was identified. Why crypto users should care Seed phrases and private keys entered into AI chats represent a catastrophic single point of failure. Anyone who obtains a seed phrase can take full control of the associated funds. If a seed phrase, private key, or other credentials were ever included in a Claude share link (or any cloud-based chat), treat them as compromised. What to do now 1) Revoke access going forward: In Claude, open Settings → Privacy → Shared Chats and revoke any shares. This can prevent future access, but it will not remove copies already saved elsewhere. 2) If a wallet seed phrase or private key was exposed: Move funds immediately to a new wallet created with a fresh seed, ideally using a hardware wallet. Assume any pasted credentials are compromised. 3) For teams: Audit shared artifacts (spreadsheets, exports, dashboards) and rotate any exposed credentials, API keys, or passwords. Remove or rehost sensitive documents that were shared via Claude. Best practices Do not paste seed phrases or private keys into cloud services. Use hardware wallets and offline key storage. Treat third-party share links as potentially public unless you have strong proof otherwise. Context This is not unprecedented. Roughly 11 months earlier, OpenAI faced a similar issue when a "Make this chat discoverable" setting led to thousands of chats being indexed, with some later captured by the Internet Archive before cleanup. In Claude's case, the trigger was not a user setting but a missing HTML directive—a reminder that minor configuration gaps can expose sensitive data at scale. Bottom line Anthropic appears to have addressed Google indexing by adding noindex tags, but third-party archives may still host copies and some search engines continue to surface share links. Anyone who has ever shared crypto credentials in Claude or similar tools should revoke shares, rotate keys, and assume compromise.