Internet Explorer 11 is not supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Massachusetts Lawmaker Uses ChatGPT to Draft Bill Limiting Its Use

State Sen. Barry Finegold used the artificial technology software to draft a bill that would regulate generative AI models and would require companies to obtain “informed consent” from individuals before using, collecting or disclosing their data.

(TNS) — A Massachusetts state lawmaker used a new chatbot that has gained popularity in recent months for its ability to write complex content to author new legislation that regulates similar programs, arguing legislators need to set guardrails on the technology while it is still in its infancy.

Sen. Barry Finegold, an Andover Democrat, used ChatGPT to help draft a bill that regulates generative artificial intelligence models like the one created by OpenAI. The proposal, filed last week, appears to be one of the first bills in Massachusetts written using the program.

In an interview Thursday morning, Finegold said he and his staff had to “nudge” the program with multiple prompts to produce a workable draft of a bill. The reason behind the bill, he said, is to show the power of the technology and the need to regulate it.

He likens ChatGPT to Facebook, where 15 years ago, he said people did not realize how influential and controversial the social media platform would become.

“Now, where we failed with Facebook in government is we did not provide proper guardrails for Facebook,” he told MassLive. “And I think what we really want to do with this legislation is to provide proper guardrails for how we deal with artificial intelligence.”

At the core of the bill are requirements that large-scale generative artificial intelligence models — described as a machine learning model with the capacity of at least one billion parameters that generates text or other forms of output — adhere to a number of operating standards.

That includes a requirement that the program not engage in discrimination or bias against any individual or group based on federally protected characteristics, prevent plagiarism by generating all text with a distinctive watermark, and implement “reasonable” security measures to protect the data of individuals used to train the model.

New York City recently banned access to ChatGPT on devices or internet networks run by the city’s education department, citing “negative impacts on student learning, and concerns regarding the safety and accuracy of content,” Chalkbeat reported.

“The most important thing is to have a watermark to show people that this is not generated from a human, that is actually generated from artificial intelligence,” Finegold said. “And I think if we had that, the New York public schools would not have banned [Chat]GPT.”

OpenAI, the company behind ChatGPT, did not immediately respond to a request for comment Thursday morning.

Finegold’s proposal also requires companies to obtain “informed consent” from individuals before collecting, using, or disclosing their data, and a requirement to “delete or de-identify” data from individuals that are no longer needed.

“The company shall conduct regular risk assessments to identify, assess and mitigate reasonably foreseeable risks and cognizable harms related to their products and services, including in the design, development, and implementation of such products and services,” the bill text said.

Any company that operates a large-scale generative artificial intelligence model like ChatGPT would have to register with the state attorney general’s office within 90 days of the bill’s effective date.

Finegold said the goal of the bill is not to make ChatGPT and similar programs “feel like, oh, they’re the enemy.”

“They’re not,” he said. “I mean, this can do amazing things and do a lot of good. But if we can provide proper guardrails, I think we can really help this company, help other artificial intelligence companies, but also protect the consumer.”

Finegold and his staff made clear that if there were any errors in the bill text, they should look to the human authors, not the program that helped draft it.

“This act has been drafted with the help of ChatGPT and any errors or inaccuracies in the bill should not be attributed to the language model but rather to its human authors,” the final section of the bill reads.

And InstaTrac, a legislative research company, also used ChatGPT to generate a one-sentence summary of the bill.

“The bill aims to regulate generative artificial intelligence models, such as ChatGPT, in order to protect the public’s safety, privacy, and intellectual property rights by setting standards for their operation, requiring registration with the attorney general, and establishing enforcement mechanisms for violations,” the summary said.

©2023 Advance Local Media LLC. Distributed by Tribune Content Agency, LLC.