Costa Rican lawmakers wanted to highlight the need to regulate AI so they asked ChatGPT to write the proposed AI regulation law for them.
The legislators prompted ChatGPT to “think like a lawyer” and then asked it to draft a bill that would regulate AI and be in line with the country’s constitution. The resulting output was submitted as-is to the legislature.
“We have had many positive reactions and many people who thought it to be very risky,” said Congresswoman Vanessa Castro, who led the introduction of the bill. Whether the perceived risk lay in the quality of laws that could be passed or the job security of lawmakers was not explained.
ChatGPT had some boilerplate suggestions like setting up a government institution to regulate AI systems governed by the 4 principles of AI ethics. Some lawmakers criticized the bill as a list of “nice to haves” rather than actionable laws.
The AI-generated bill suggested, “We should regulate based on fundamental rights and international conventions,”. In defense of ChatGPT, the rest of the world can’t agree on what those rights and conventions should be and how to ensure them either.
The draft bill demonstrated the potential contribution AI can make to the legislative process. But it also highlighted how important human discussions are to ensure that social issues aren’t sacrificed for efficiency and cold machine logic.
AI regulation in Latin America makes slow progress
Costa Rica is the eighth country in Latin America to pass or discuss laws related to the regulation of AI in the last year. As technology embraces AI, legislators are racing to keep up with the ethical issues the new technologies are creating.
Brazil legislators have been discussing potential AI laws for the last 4 years. In May Brazilian Senate Speaker Rodrigo Pacheco finally submitted a bill outlining how AI should be regulated to protect Brazilian citizens.
Key aspects of the bill focus on ensuring that issues surrounding inequalities and racism aren’t reinforced with the introduction of AI. These challenges were unresolved in Brazil before AI so the potential for risks associated with social credit scores, access to finance, and public surveillance being exacerbated by AI are real.
The bill is based on panel discussions that started back in 2021, with some experts saying that developments since then have made the bill dated before it has been written into law.
For legislators to keep up with the pace of AI technology they may need to rely on using ChatGPT more often and not just for the sake of novelty.