AI Exec: We May Have Pulled Off "The Largest Theft of Labor in Human History"
6 hours ago
- AI companies like OpenAI and Microsoft are accused of massive copyright theft by training models on copyrighted content from newsrooms without permission or compensation.
- Internal documents reveal executives knew they were destroying the economic base for publishers and reducing traffic to news sites by up to 90% with AI search.
- The companies acknowledged a 'doom loop' where AI threatens the content supply chain and degrades the internet, yet they continue despite knowing the harm.
- Greg Brockman, OpenAI president, celebrated the AI's ability to reproduce news articles and approved hacks to bypass paywalls for scraping content.
- Training data was deliberately sourced from high-quality content, including paywalled news articles, often in violation of terms of service.
- The companies stripped copyright notices and ownership data from datasets, showing a systematic disregard for creators' rights.
- After lawsuits were filed, OpenAI filtered out outputs from suing publishers, not to prevent infringement but to hinder evidence collection.
- OpenAI found over 400,000 25-word overlaps between ChatGPT responses and Mother Jones articles, indicating significant regurgitation.
- The legal brief argues AI companies are trapped in a prisoner's dilemma, where individual incentives to keep taking content free outweigh collective benefits of paying.
- The article concludes that AI executives cannot be trusted to make ethical decisions, as they knowingly make harmful choices for short-term gain.