Microsoft and OpenAI were sued Wednesday by sixteen pseudonymous individuals who allege the companies’ ChatGPT-based AI products collected and shared their personal information without proper notice or consent.
The complaint [PDF]The lawsuit, filed in federal court in San Francisco, California, alleges that the two companies ignored legal means of collecting data for their AI models and chose to collect it without paying for it.
“Despite established protocols for the acquisition and use of personally identifiable information, defendants took a different approach: theft,” the lawsuit reads. “They have systematically scraped 300 billion words from the internet, ‘books, articles, websites and posts – including personal data obtained without consent’.” OpenAI did this in secret and without registering as a data broker, as it is claimed applicable law would be required.”
It is alleged that the two companies use their AI products to “collect, store, track, share and disclose” millions of people’s personal information, including product details, account information, names, contact details, login information, email and payment information, transaction records , browsing data, social media information, chat logs, usage data, analytics, cookies, searches and other online activity.
The complaint alleges that Microsoft and OpenAI have embedded millions of people’s personal data into their AI products that reveal hobbies, religious beliefs, political views, voting results, social and support group memberships, sexual orientation and gender identity, work histories, family photos, etc. reflect. Friends and other data from online interactions.
OpenAI has developed a family of text-generating large language models that includes GPT-2, GPT-4, and ChatGPT; Not only is Microsoft committed to the technology, but it has also integrated it into all areas of its empire, from Windows to Azure.
“Regarding personally identifiable information, the defendants fail to adequately filter it from training models, putting millions of people at risk of having that information promptly or otherwise disclosed to strangers around the world,” the lawsuit reads citing the March 18 notice from The Register. Special report 2021 on this topic.
The 157-page complaint is replete with media and academic citations expressing concerns about AI models and ethics, but little detail on specific claims.
For the 16 plaintiffs, the complaint shows that they used ChatGPT, as well as other internet services such as Reddit, and expected that their digital interactions would not be integrated into an AI model.
It remains to be seen how, if at all, content and metadata created by the plaintiff was actually exploited and whether ChatGPT or other models will reproduce this data.
OpenAI has historically dealt with the reproduction of personal information by filtering it.
The lawsuit seeks certification of a class action and damages of $3 billion — though that number is believed to be a placeholder. The actual damages would be determined if the plaintiffs prevail, based on the court’s findings.
The complaint alleges that Microsoft and OpenAI violated America’s Electronic Privacy Communications Act by acquiring and using private information and unlawfully intercepting communications between users and third-party services via integrations with ChatGPT and similar products.
The Sueball further alleges that the defendants violated the Computer Fraud and Abuse Act by intercepting interaction data via plugins.
It also alleges violations of the California Invasion of Privacy Act and the Unfair Competition Act, the Illinois Biometric Information Privacy Act and the Consumer Fraud and Fraudulent Business Practices Act, and New York City Business Law, and various general tort claims such as negligence and unjust enrichment.
Microsoft and OpenAI declined to comment.
Microsoft, its GitHub subsidiary and OpenAI were sued last November for allegedly reproducing the code of millions of software developers in violation of licensing requirements via GitHub’s Copilot service, which is based on an OpenAI model. This case is not yet closed. ®