Ming Min originated from the temple of Wofei
Qubits | Official account QbitAI
With two lawsuits a day, ChatGPT is now in big trouble.
The reasons are all related to data.
First, 16 people anonymously sued OpenAI and Microsoft, arguing that they used and leaked personal privacy data without permission, and claimed up to $3 billion.
Immediately afterwards, the two full-time authors argued that OpenAI used their novel training ChatGPT without permission, which constituted infringement.
And the plaintiffs on both sides say that OpenAI has scavenged too much data from the Internet, and that there are many other people's data being used illegally.
There is a lot of discussion online, and some people say that writers should defend their rights and be properly paid.
But others argue that generative AI learning about internet data is like a student visiting a work of art and should be open.
Both lawsuits are now in federal court in San Francisco, California.
The defendant, OpenAI, did not respond publicly for the time being.
16 people claimed 3 billion
Let's start with the first lawsuit.
On June 28, 16 anonymous individuals filed a lawsuit against OpenAI and Microsoft, arguing that ChatGPT-based AI products collected and leaked their personal information without informing or obtaining their consent.
This behavior violates legal policies related to AI model data, and there is no payment.
They scraped 300 billion words from the internet, books, articles, websites, and posts, including personal information obtained without permission.
This includes account information, name, contact details, email, payment information, transaction history, browser data, social media information, chat data, cookies, etc.
This allows personal information to be embedded in their AI products, which can reflect personal preferences, opinions, work history and even family photos.
The plaintiffs argue that OpenAI's failure to adequately filter this sensitive information puts millions of people at risk of information being compromised.
And they allege that OpenAI was "secretive" and did not register as a data broker as required by applicable law.
Clarkson, the law firm in charge of the lawsuit, has previously led massive class-action lawsuits over issues such as data breaches and false advertising.
But it is uncertain whether the case will be won.
Katherine Gardner, an intellectual property lawyer, said that when users upload content to social platforms or other websites, they grant the platform a broad license to use their content.
Therefore, it is relatively challenging for ordinary users to get compensation for the data acquisition of AI models.
It has also been found that the 157-page lawsuit quotes a lot of media and academic discussions and warnings on AI ethics, but there are not many specific cases that have actual impact.
On the other hand, two American authors also sued OpenAI on the same day, arguing that their work was taken to train ChatGPT.
Concrete evidence is that ChatGPT produces "very accurate" summaries of their books, which is enough to show that the books are included in the database.
Paul Tremblay and Mona Awad say ChatGPT is infringing on the authors' copyrights by copying data from thousands of books without permission.
The indictment estimates that OpenAI's training data contains at least 300,000 books, many of them from infringing websites.
For example, when OpenAI disclosed GPT-3 training data, it said that it contained two Internet book corpora, accounting for about 15%. The authors of the lawsuit believe the data came from shadow library websites such as Library Genesis and Sci-Hub.
And in 2018, OpenAI revealed that the data they fed GPT-1 included 7,000+ novels. The prosecution argues that the books were copied by researchers from some pirated websites without the authors' consent or authorization.
There has been no public response from OpenAI.
In fact, since the fire of generative AI, it is really not uncommon for AI companies to suffer copyright and data privacy lawsuits.
OpenAI has been mired in a data turmoil
Back last November, OpenAI and GitHub's code assistant Copilot was sued by programmers.
The plaintiffs argued that Copilot had infringed the copyrights of many original code authors while also leaking user privacy, so they filed a lawsuit in court for $9 billion.
But so far, the case has not been finalized.
On the AI painting side, Stability AI, Midjourney, etc. are also riddled with lawsuits.
Discovering that you could generate a personal style of painting by typing your own name, painter Kelly McKernan joined two other creators in a lawsuit.
The team of lawyers is the same as the one sued Copilot above.
And the website platform side is also unwilling to be whitely prostituted data.
This is not the reason Musk recently limited Twitter streaming, and the reason given is that "I don't want AI to prostitute data." At present, the restrictions are still in place, and it is not yet known how they will change after that.
Reddit said that the data on its own platform is very valuable, so it launched a paid API, the price is not low, this operation has led to many third-party applications forced to close.
In short, after AI large-scale data grabbing training on the network, privacy security issues and copyright issues have emerged one after another, and more and more people have been involved.
However, at present, how to calculate the copyright issue, the industry has not yet formed a definite norm.
Only the Japanese side has reported that it will not implement copyright protection for the data used to train the AI, because it can accelerate the development of AI.
Reference Links:
[1]https://www.theregister.com/2023/06/28/microsoft_openai_sued_privacy/
[2]https://www.reuters.com/legal/lawsuit-says-openai-violated-us-authors-copyrights-train-ai-chatbot-2023-06-29/
[3]https://www.washingtonpost.com/technology/2023/06/28/openai-chatgpt-lawsuit-class-action/
— End —
Qubits QbitAI · Headline number signed
Follow us and be the first to know the latest scientific and technological trends