Freelance journalist Conrad Quilty-Harper once made what he now calls a stupid decision. He upgraded his personal ChatGPT account to a Business plan. The prompts from OpenAI made it sound like a smart move. Access to advanced features. Better controls. Yet something basic vanished.
No easy way to export his chat history. The standard data export button that Free, Plus and Pro users enjoy simply does not exist for Business and Enterprise workspaces. Quilty-Harper took matters into his own hands. He built a free tool. It scrapes what OpenAI won’t hand over. And he published it on GitHub this week.
The Register first broke the story of this workaround. Its article details how scrapemychats lets users pull conversations, files and attachments into a local, searchable HTML archive. The tool runs through a visible Chrome browser session. It mimics human behavior to dodge bot protections. Slow by design. A collection of 600 chats can take hours. But it works. For now.
OpenAI’s own help documentation confirms the restriction. Its page on exporting ChatGPT history states clearly that the feature is not available for ChatGPT Business or Enterprise workspaces. Free, Plus, Pro and some Education customers get the option. Business users do not. The company offered no explanation when The Register asked. Silence followed.
Enterprise admins have one alternative. OpenAI’s Compliance Platform gives them access to workspace conversation logs. Yet those logs last only 30 days unless archived elsewhere. The platform targets compliance teams. It integrates with eDiscovery and SIEM tools from partners like CrowdStrike and Microsoft. Business subscribers lack even that. They sit in a data no-man’s land.
Quilty-Harper voiced his frustration in an email to The Register. “I made the stupid decision to upgrade from a personal account to a Business one a while back as they prompted, and for some reason OpenAI doesn’t let you export the data easily.” He suspects lock-in tactics. Keep customers tethered. Make leaving painful. “I would quite like to know the reason why ChatGPT doesn’t make it easy for paying customers to get access to their data,” he added. “I have my suspicions.”
The tool itself operates with care. It never leaves the user’s machine. No passwords shared. No data sent to third parties. It captures JSON responses from ChatGPT’s own frontend calls. Downloads attachments through the files API. Saves generated files from the code interpreter. The result? A clean folder structure. Individual conversation JSON and Markdown files. A single viewer.html with instant full-text search. Keyboard navigation. Offline access. Even from a USB drive.
Limitations exist. Some older uploaded files disappear from OpenAI’s servers after a retention period. The tool logs those failures but cannot recover them. Headless automation fails because of bot detection. The script paces requests to avoid throttling. Interruptions? It resumes where it left off. Smart design from someone who needed a practical fix.
Quilty-Harper anticipated pushback. “I expect OpenAI will stop this from working shortly after you write about it,” he told The Register. “It’s probably a breach of their terms of service. But hey, their entire business model is based around scraping copyrighted material. This just allows you to get the content you wrote or pasted or created (and generated) back.”
His point lands with force. OpenAI built its empire on vast troves of web data. Training models on books, articles, code. Copyright battles continue in courts worldwide. Yet when users want their own generated conversations? Friction appears. Especially at the paid enterprise level where data stakes run higher.
Community forums echo the anger. A long thread on OpenAI’s community site from early 2025 labels the missing export a potential GDPR violation. Users who merged personal chats into Team or Business workspaces lost access forever. Merges are permanent. Exports unavailable. Support responses loop back to personal account instructions that do not apply. Trust erodes.
Recent searches turn up similar complaints. A Medium post from this year explains why ChatGPT’s built-in export falls short for professionals. It lacks import options. Formatting breaks in some tools. Business users get nothing at all. Another article on export changes notes OpenAI quietly altered file structures in April 2026. Blank conversations appeared. Third-party readers scrambled to adapt.
Privacy shifts add context. A court ruling last year freed OpenAI from storing all deleted chats indefinitely. The company now deletes temporary conversations after 30 days for most users. Good for privacy in one sense. Yet it sharpens the contrast. When OpenAI wants to purge data it acts fast. When customers seek to preserve their own work? Barriers remain.
So why the policy? Lock-in offers one theory. Once teams build knowledge bases inside ChatGPT workspaces, switching providers means losing context. Prompts. Analyses. Code snippets. Institutional memory. All trapped. Competitors like Anthropic or Google offer their own chat tools. Data portability could accelerate churn. OpenAI appears unwilling to ease that path for its highest-paying customers.
Enterprise deals emphasize control. Admins manage sharing. Data stays within the organization. Yet individual contributors or freelancers who upgrade personal accounts lose personal ownership. Quilty-Harper’s experience highlights that trap. Upgrade prompted. Export removed. No warning about irreversible data migration.
The scrapemychats README lays out setup for non-technical users. Install Python. Grab Chrome. Run a test with three chats. Verify. Then full export. Build the viewer. Optional categorization by keywords for work and personal separation. AI assistants can even walk users through it step by step. The author collaborated with Claude on parts of the code. Irony noted.
Installation takes 20 minutes for beginners. The export itself demands patience. Throttling hits hard. Each chat loads fully in the browser. Screenshots, tables, images captured. Everything saved locally. The viewer displays conversations like an email client. Categories on the left. List in the middle. Full view on the right. Thumbnails. Search that highlights terms in place.
OpenAI’s terms of service likely prohibit automation. The disclaimer on the GitHub repo acknowledges the risk. Use at your own peril. For your own data. The tool behaves like a patient human. Slow clicks. Waits. No aggressive scraping. Still. Companies have shut down similar projects before.
But. This one feels different. It addresses a genuine gap in data rights. Users create the content. They paste sensitive documents. Generate reports. Brainstorm strategies. That output holds value. Losing access upon cancellation or workspace closure borders on unfair. Especially when the same company scrapes the open web without hesitation.
Regulators watch closely. Europe’s GDPR demands data portability. California’s CCPA includes similar rights. Enterprise customers negotiate custom contracts. Yet the help pages reveal a one-size-fits-most approach that leaves Business users exposed. No export. Short log retention. Complex compliance tools reserved for bigger spenders.
Quilty-Harper released version 1.0 just hours before The Register story. It already carries security fixes. The project sits at 17 stars on GitHub. Small but targeted. One fork. Contributions from the original author and an AI. Practical utility over hype.
Industry insiders see this as another data sovereignty moment. Organizations pour millions into AI platforms. They feed proprietary information into chats. Prompts become intellectual property. Outputs shape decisions. Yet control over that history rests with the vendor. Export tools like this one reclaim some agency. Until patched.
What happens next? OpenAI could close the loophole. Update bot protections. Rate limits. Or better yet, add official export for Business users. The latter would build goodwill. Reduce reliance on unofficial scrapers. Align with privacy rhetoric the company often promotes.
Until then tools like scrapemychats fill the void. They remind everyone that data created on a platform should not vanish when the relationship ends. Users write it. They paste it. They generate it with paid compute. Ownership questions linger. The journalist’s simple script makes one thing clear. If the company won’t provide an exit, clever individuals will build their own door.
WebProNews is an iEntry Publication