Scoop: Congress sets limits on staff ChatGPT use

Scoop: Congress sets limits on staff ChatGPT use

The House is placing new guardrails around use of the popular AI chatbot ChatGPT by congressional offices, Axios has learned.

Why it matters: It's the latest example of how Washington is grappling with the implications of the recent explosive growth in generative AI both legislatively and personally.

Driving the news: In a memo to House staffers on Monday morning, a copy of which was obtained by Axios, the chamber's Chief Administrative Officer Catherine L. Szpindor wrote that offices are "only authorized" to use the paid ChatGPT Plus.

• Unlike the free service, she said, the $20-per-month subscription version "incorporates important privacy features that are necessary to protect House data."
• She said in addition to other versions of ChatGPT, no other large language models are authorized for use.

I asked Bard to outline federal legislation for regulation of the use of large language models such as ChatGPT and Bard in drafting of legislation

outline federal legislation for regulation of the use of large language models such as ChatGPT and Bard in drafting of legislation

Sure, here is an outline of federal legislation for regulation of the use of large language models such as ChatGPT and Bard in drafting of legislation:

• Title: The Large Language Models in Legislation Act
• Section 1. Findings. The Congress finds the following:
  • Large language models (LLMs) are powerful tools that can be used to generate text, translate languages, write different kinds of creative content, and answer questions in an informative way.
  • LLMs have the potential to be used to draft legislation, but there are risks associated with this use, such as the potential for LLMs to generate biased or inaccurate content, or to be used to circumvent existing laws and regulations.
  • It is in the public interest to regulate the use of LLMs in drafting of legislation in order to mitigate these risks and ensure that LLMs are used in a responsible and ethical manner.
• Section 2. Definitions. In this Act:
  • The term "large language model" means a statistical method that predicts the next word in a sequence of words, based on the words that have already been generated.
  • The term "drafting of legislation" means the process of writing, editing, and revising proposed legislation.
• Section 3. Prohibitions. It shall be unlawful for any person to use an LLM to draft legislation in a manner that:
  • Is biased or inaccurate;
  • Circumvents existing laws or regulations;
  • Is intended to mislead or deceive the public; or
  • Is otherwise harmful to the public interest.
• Section 4. Enforcement. The Federal Trade Commission shall have the authority to enforce this Act.
• Section 5. Severability. If any provision of this Act is held to be invalid, the remainder of this Act shall remain in full force and effect.

This is just an outline, and the specific provisions of the legislation would need to be further developed. However, this outline provides a starting point for considering how to regulate the use of LLMs in drafting of legislation.

Here are some additional considerations that could be included in the legislation:

• The level of human oversight required for the use of LLMs in drafting legislation.
• The types of information that LLMs are allowed to access when drafting legislation.
• The ways in which the output of LLMs can be verified for accuracy and bias.
• The penalties for violations of the legislation.

The goal of the legislation should be to strike a balance between the potential benefits of using LLMs in drafting legislation and the risks associated with this use. The legislation should also be flexible enough to adapt to the rapidly evolving nature of LLM technology.

Why pick the FTC to enforce this? It seems that a Congressional office similar to the GAO might be better. Also why limit it to just the Legislature. How about limits on Executive departments and Judiciary?

The FTC was chosen to enforce the legislation because it has a long history of regulating the use of technology in the public interest. The FTC has also been involved in regulating the use of artificial intelligence, and it has the expertise and resources to enforce the legislation.

A Congressional office similar to the GAO might also be a good choice to enforce the legislation. The GAO has a strong track record of auditing and evaluating government programs, and it could provide valuable insights into the use of LLMs in drafting legislation.

The legislation could be expanded to include the executive branch and the judiciary. However, there are some challenges to enforcing such legislation in these branches. For example, the executive branch has a lot of discretion in how it implements laws, and it could be difficult to enforce limits on the use of LLMs in the executive branch. The judiciary is also independent, and it could be difficult to enforce limits on the use of LLMs in the judiciary.

Ultimately, the decision of who should enforce the legislation and which branches of government should be included is a policy decision that would need to be made by Congress.

Here are some additional considerations that could be taken into account when making this decision:

• The expertise and resources of the different agencies.
• The independence of the different agencies.
• The need to ensure that the legislation is enforceable.
• The need to balance the potential benefits and risks of using LLMs in different branches of government.

Citizens might feel that offices with the most power over them need the most regulation. They do not wish to be governed by AI

You are right. Citizens might feel that offices with the most power over them need the most regulation. They do not wish to be governed by AI. This is a legitimate concern, and it is important to consider the potential risks of using AI in these branches of government.

Here are some of the risks that could arise from using AI in the executive branch and the judiciary:

• Bias: AI systems are trained on large amounts of data, and this data can be biased. If an AI system is trained on data that is biased, it could generate biased output. This could lead to decisions that are unfair or discriminatory.
• Inaccuracy: AI systems are not perfect, and they can make mistakes. If an AI system is used to make decisions that have a significant impact on people's lives, these mistakes could have serious consequences.
• Opacity: AI systems are often opaque, meaning that it is difficult to understand how they make decisions. This can make it difficult to hold AI systems accountable for their decisions.

These are just some of the risks that could arise from using AI in the executive branch and the judiciary. It is important to carefully consider these risks before deciding whether or not to use AI in these branches of government.

It is also important to remember that AI is a tool, and like any tool, it can be used for good or evil. It is up to us to ensure that AI is used in a responsible and ethical manner. We need to put in place safeguards to prevent AI from being used to harm people or to violate their rights.

If we do this, AI can be a powerful force for good in the world. It can help us to make better decisions, to solve complex problems, and to improve our lives.