OpenAI hit with class-action lawsuit over web-scraping data to train ChatGPT

Every forum post, late-night comment, or digital footprint you left behind might already be inside ChatGPT. But did OpenAI ever ask for your permission before turning it into a commercial goldmine?

California-based Clarkson Law Firm just filed a massive class-action lawsuit in the Northern District of California against OpenAI and its key backer, Microsoft. The suit, brought on behalf of 16 individuals, accuses the tech giants of scraping around 300 billion words from the web without consent, stealing copyrighted material and the sensitive personal information of millions of people.

Medical records, logins, and a $3B price tag

According to the complaint, OpenAI did not just vacuum up Wikipedia entries and public news articles. The scraped data allegedly included names, login credentials, payment details, and even sensitive medical records belonging to everyday Americans.

The lawsuit pulls no punches, describing the data collection as outright theft and noting that OpenAI never registered as a data broker as required by law. The allegations also extend beyond web scrapers: plaintiffs claim OpenAI unlawfully pulled user data directly through ChatGPT and through integrations with third-party apps like Snapchat and Stripe.

The plaintiffs are asking for $3 billion in damages, though legal observers note this figure is likely a placeholder sum. Beyond cash, the lawsuit demands a temporary freeze on the commercial rollout and development of ChatGPT until strict safeguards are implemented, alongside an independent "AI council" to oversee the tech.

The crumbling defense of "fair use"

Ryan Clarkson, managing partner at the firm, made it clear why they went after Sam Altman's company first: OpenAI sparked this entire generative AI arms race, making them the natural primary target.

AI labs have historically shielded their massive scraping sprees behind copyright law's "fair use" doctrine, arguing that training algorithms creates a transformative work. But intellectual property lawyer Katherine Gardner points out that while artists with copyrighted works might have a fighting chance in court, everyday internet users will struggle to win damages. Most users agreed to sweeping terms of service the moment they signed up for social media platforms.

Pushback is mounting everywhere. Universal Music Group asked Apple and Spotify to block AI scrapers in April. Reddit is closing off its free data spigot, and Elon Musk previously threatened legal action over Microsoft accessing Twitter data.

Why it matters

For regular internet users, this lawsuit validates a growing unease about being unpaid fuel for generative AI, but you should not expect a settlement check in the mail anytime soon. Proving tangible financial harm from a casual blog post or forum comment is legally brutal, especially against platforms with airtight terms of service. For now, the most practical takeaway is self-defense: stop sharing private data in chat prompts.

For the AI industry, the era of treating the public internet as an all-you-can-eat buffet appears to be ending. If the court forces OpenAI to turn over details on what actually went into GPT-4, or if regulators step in, the cost of acquiring legitimate training data will skyrocket.

We think combining privacy violations with copyright infringement is a clever tactical move by the plaintiffs. Fighting solely on copyright gives OpenAI room to hide behind fair use doctrines. Bringing in sensitive personal records, passwords, and medical data, however, exposes OpenAI to stringent federal wiretapping and state privacy laws. That said, demanding an outright halt to ChatGPT's commercialization looks far too extreme to survive judicial scrutiny.