110,000 ChatGPT Conversations Still Live on Archive.org Despite OpenAI Cleanup

OpenAI may have removed nearly 50,000 shared ChatGPT conversations from Google’s search index, but the cleanup is far from complete. A new investigation by Digital Digging, led by researcher Henk van Ess and Belgian collaborator Nicolas Deleur, reveals over 110,000 ChatGPT chats still accessible via the Wayback Machine on Archive.org.

These conversations, originally shared by users through the “Share” button on ChatGPT, were intended as temporary links. Unbeknownst to many, these links were permanently archived, including deeply personal, controversial, and in some cases, potentially illegal content.

Archive.org’s Mark Graham confirmed that no large-scale exclusion request has been made by OpenAI regarding chatgpt.com/share URLs, meaning the data remains online and searchable—just not by Google.

The investigation unearthed alarming content, including:

• An Italian lawyer revealing a corporate plan to displace indigenous Amazonian communities for a hydroelectric project.
• A user in an Arabic-speaking country using ChatGPT to write political dissent targeting their authoritarian regime.
• Dozens of Persian-language chats documenting academic fraud, where students used ChatGPT to generate entire research papers and submitted them for grades.

While most shared chats were harmless, others expose serious ethical and legal breaches. And despite OpenAI’s swift action to remove indexed URLs, the internet’s golden rule still holds: once online, always online.

“We’re not sharing direct links,” van Ess noted. “This is a warning, not a roadmap for exploitation.”

Grok’s shared conversations are also briefly visible via Google but are being removed. OpenAI has not responded to requests for comment.