Brazilian city enacts an ordinance secretly written by a surprising new staffer: ChatGPT

City lawmakers in Brazil have enacted what appears to be the nation’s first legislation written entirely by artificial intelligence — even if they didn’t know it at the time.

The experimental ordinance was passed in October in the southern city of Porto Alegre and city councilman Ramiro Rosario revealed this week that it was written by a chatbot, sparking objections and raising questions about the role of artificial intelligence in public policy.

Rosario told The Associated Press that he asked OpenAI’s chatbot ChatGPT to craft a proposal to prevent the city from charging taxpayers to replace water consumption meters if they are stolen. He then presented it to his 35 peers on the council without making a single change or even letting them know about its unprecedented origin.

“If I had revealed it before, the proposal certainly wouldn’t even have been taken to a vote,” Rosario told the AP by phone on Thursday. The 36-member council approved it unanimously and the ordinance went into effect on Nov. 23.

“It would be unfair to the population to run the risk of the project not being approved simply because it was written by artificial intelligence,” he added.

The arrival of ChatGPT on the marketplace just a year ago has sparked a global debate on the impacts of potentially revolutionary AI-powered chatbots. While some see it as a promising tool, it has also caused concerns and anxiety about the unintended or undesired impacts of a machine handling tasks currently performed by humans.

Porto Alegre, with a population of 1.3 million, is the second-largest city in Brazil’s south. The city’s council president, Hamilton Sossmeier, found out that Rosario had enlisted ChatGPT to write the proposal when the councilman bragged about the achievement on social media on Wednesday. Sossmeier initially told local media he thought it was a “dangerous precedent.”

The AI large language models that power chatbots like ChatGPT work by repeatedly trying to guess the next word in a sentence and are prone to making up false information, a phenomenon sometimes called hallucination.

All chatbots sometimes introduce false information when summarizing a document, ranging from about 3% of the time for the most advanced GPT model to a rate of about 27% for one of Google’s models, according to recently published research by the tech company Vectara.

In an article published on the website of Harvard Law School’s Center of Legal Profession earlier this year, Andrew Perlman, dean at Suffolk University Law School, wrote that ChatGPT “may portend an even more momentous shift than the advent of the internet,” but also warned of its potential shortcomings.

“It may not always be able to account for the nuances and complexities of the law. Because ChatGPT is a machine learning system, it may not have the same level of understanding and judgment as a human lawyer when it comes to interpreting legal principles and precedent. This could lead to problems in situations where a more in-depth legal analysis is required,” Perlman wrote.

Porto Alegre’s Rosario wasn’t the first lawmaker in the world to test ChatGPT’s abilities. Others have done so in a more limited capacity or with less successful outcomes.

In Massachusetts, Democratic state Sen. Barry Finegold turned to ChatGPT to help write a bill aimed at regulating artificial intelligence models, including ChatGPT. Filed earlier this year, it has yet to be voted on.

Finegold said by phone on Wednesday that ChatGPT can help with some of the more tedious elements of the lawmaking process, including correctly and quickly searching and citing laws already on the books. However, it is critical that everyone knows ChatGPT or a similar tool was used in the process, he added.

“We want work that is ChatGPT generated to be watermarked,” he said, adding that the use of artificial intelligence to help draft new laws is inevitable. “I’m in favor of people using ChatGPT to write bills as long as it’s clear.”

There was no such transparency for Rosario’s proposal in Porto Alegre. Sossmeier said Rosario did not inform fellow council members that ChatGPT had written the proposal.

Keeping the proposal’s origin secret was intentional. Rosario told the AP his objective was not just to resolve a local issue, but also to spark a debate. He said he entered a 49-word prompt into ChatGPT and it returned the full draft proposal within seconds, including justifications.

“I am convinced that … humanity will experience a new technological revolution,” he said. “All the tools we have developed as a civilization can be used for evil and good. That’s why we have to show how it can be used for good.”

And the council president, who initially decried the method, already appears to have been swayed.

“I changed my mind,” Sossmeier said. “I started to read more in depth and saw that, unfortunately or fortunately, this is going to be a trend.”

———-

Savarese reported from Sao Paulo. AP journalists Steve LeBlanc in Boston and Matt O’Brien in Providence, Rhode Island, contributed to this report.