ChatGPT is accused of plagiarism and defamation; understand cases – 08/30/2023 – Tech

Brazilian courts have not defined when the use of ChatGPT can be a case of Justice. There are only six precedents on the JusBrasil website that cite the text-generating artificial intelligence created by OpenAI.

In the case that had the greatest impact, the minister of the TSE (Superior Electoral Court) Bento Gonçalves fined a lawyer for using the tool to make a petition.

The platform is also cited as a “challenge for the world of work” in Labor Court decisions and as a negative example for the vagueness of arguments in four petitions signed by lawyers.

Brazilian judicial authorities have not addressed accusations of plagiarism, defamation and violation of privacy for which OpenAI already responds in other countries.

In Brazil, there is no clear definition even of how AI can be used in the courts. The same lawyer fined by Minister Gonçalves, Fábio de Oliveira Ribeiro, has been appealing to the CNJ (National Council of Justice) since January to prevent judges from using ChatGPT to issue or substantiate decisions.

The request for advance relief made by the lawyer was denied the following month, as there was no evidence of such use. The case awaits trial.

Lawsuits against text-generating AI around the world also remain unsolved, as legal experts debate who should be held accountable for possible errors in new AI systems and how best to do so. There are cases, for example, in which ChatGPT “hallucinates” and invents slander about people.

The report separated episodes to explain who is responsible for the errors of artificial intelligence, since the technology does not have legal personality to be taken to court. Target of the lawsuits, OpenAI did not respond to requests for comment on the report.

CALUMNY AND DEFAMATION CHARGES

In Brazil, there are no defamation lawsuits against OpenAI or another artificial intelligence platform being processed without secrecy of justice, according to the JusBrasil platform.

At the request of the report, the TJSP (Court of Justice of São Paulo) also consulted, without results, certificates in the São Paulo Court citing the startup.

In the US, a Georgia broadcaster sued OpenAI for libel, with allegations that ChatGPT fabricated a lawsuit with false accusations of misappropriation of funds and manipulation of financial records while he was an executive of an organization. The radio professional, in fact, never had any dealings with such a company.

In a document filed with the court asking for the action to be dismissed, OpenAI said that “there is near universal agreement that the responsible use of AI includes verifying the veracity of information presented by the AI ​​before using or sharing it”.

OpenAI declined to comment on specific cases. However, in a note to the Washington Post, he said that he is transparent during the registration process for new users and that improving the factual accuracy of language models is one of the main focuses of the company. “We are making progress.”

An American law professor and an Australian mayor went through similar situations.

According to University of Minnesota professor Alan Rozenshtein, anyone who feels defamed by ChatGPT should file a lawsuit against OpenAI for making a product available that systematically attributes false claims to third parties.

“It’s a complicated lawsuit that needs to show the company’s negligence in technology development,” he told Sheet.

INFRINGEMENT OF INTELLECTUAL PROPERTY

A group of artists from California sued the American Justice to ask OpenAI for reparation, under the allegation that their styles were copied by Dall-E —artificial intelligence that generates images.

Artists want to be approached to authorize or refuse their works to be used in the training of these technologies.

In the same direction, the Wall Street Journal and the CNN TV network published articles stating that OpenAI used texts from its websites and 18 other vehicles to train ChatGPT —the information would have been revealed by the AI ​​itself, in response to a question .

The ChatGPT algorithm was fed with more than 45 terabytes of texts, collected from the internet without the authors’ authorization. Sources are specified in the GPT-3 engine release article behind ChatGPT. In disclosing the latest version of this model, GPT-4, OpenAI did not publish the data used in development.

On August 11th, the specialized website Ars Technica showed that OpenAI added details about the operation of the scraper robot used to feed its artificial intelligence, the GPBTot. With this information, it is possible to block data mining from the creator of ChatGPT.

The intellectual property policy for training AIs varies from country to country, according to InternetLab’s culture and knowledge coordinator, Alice Lana.

Brazil has not yet defined its norms. The copyright bill pending in Congress includes remuneration for authors for content used by artificial intelligence.

“United States, Japan and Singapore and some countries of the European Union have already made legislative changes that allow data and text mining to take place in databases, without violating copyright”, says Lana.

This ensures legal certainty in the development of generative artificial intelligence. “The idea is that the content was not used by a person, but processed by machines”, says the researcher.

In the US, this exception to intellectual property law only applies to research. US officials have not decided whether AI applications developed by for-profit companies fall under this understanding.

ARE WORKS CREATED BY AI INTELLECTUAL PROPERTY?

Authorities also debate whether content generated by artificial intelligence is protected by intellectual property. There are ongoing lawsuits on this topic in the US and Australia.

Although the elaboration of texts and arts with artificial intelligence requires human participation in the production of instructions (prompt, in technical language), this creation is only possible by training with data available on the internet.

According to Lana, most theorists in the field argue that any content generated by AI should automatically enter the public domain, as with books and music after a certain time since the author’s death.

“It makes no sense to say that the owner of that work is artificial intelligence. It would be as if, at the beginning of computing, we said that a poem typed on the computer belonged to the computer”, says Lana.

The US Copyright Office ruled in March that images created by the AI ​​generator Midjouney and then used in a graphic novel could not be protected, according to a report by Reuters. This platform became known for having created the fake photo of Pope Francis in a white coat.