AI is good at understanding bureaucratic texts, which are more subject to predictable patterns than poetry or narrative, for example. Even if that bureaucracy is thousands of years old. She also seems to be good at handling laws, given that a Brazilian councilman was recently able to introduce a regulation generated by ChatGPT and get it passed without changes.

Perhaps this is why the UK has recently taken an unexpected step by incorporating artificial intelligence into its judicial system by allowing judges to use ChatGPT (and other AI tools) to assist them in drafting legal decisions and other bureaucratic tasks related to justice.

This decision has already generated, of course, a debate about the correlation of risks and benefits that would come hand in hand with implementing AI in the legal field as well.

How did this happen? Well, by publishing an 8-page guide (PDF), prepared by the Judicial Office of the United Kingdom and aimed at judges in England and Wales (Scotland and Northern Ireland are autonomous in this area). It guides them on how to use AI to perform their tasks more effectively, and explicitly authorizes them to use ChatGPT.

Chatbots In IT For Smarter Helpdesk Experience | Data Semantics

“—ChatGPT, did the defendant order the code red? —…”

Although there is a temptation to stick only with the shocking fact that “judges are going to write their sentences with ChatGPT!”, the guide also warns about the potential risks of using AI in the judicial system… such as the well-known possibility that AI-generated answers are inaccurate, incomplete, misleading, or biased, so it recommends judges verify the accuracy of the generated answers before making decisions that may affect the people involved.

The guide also highlights privacy-related concerns, noting that AI companies can collect information from users’ interactions with chatbots. Therefore, judges are urged to treat any information entered into a chatbot’s interface as if it were public.

As novel as it may seem, some judges had actually used ChatGPT before: a judge at the Court of Appeal for England and Wales, Sir Geoffrey Vos, described ChatGPT as a “very useful” tool that he used to summarize legal theories and copy them into an official ruling. However, he points out that he used the tool only to rework information with which he was already familiar, and which did not include confidential data.

The experiments, with soda (on the other side of the Atlantic)

Britain’s attitude toward the legal use of AI seems more open to experimentation than that of the most prominent of its former colonies: in the United States, a lawsuit for professional intrusion has put an end to an experiment by DoNotPay to use an artificial intelligence to act as a lawyer to generate documents for claims.

They claimed that ChatGPT “does not have a law degree”, although the truth is that it has proven to be able to pass the US bar exam.

ChatGPT has also become the possible reason for the end of the career of a lawyer with 30 years of experience for inventing court rulings. Or rather, for letting ChatGPT make them up and not personally verifying them below. Forced to justify himself before the judge, the lawyer explained that he was unaware that AIs could ‘hallucinate’ and thus generate totally invented content.

Comments

LEAVE A REPLY

Please enter your comment!
Please enter your name here