How I used Anthropic Claude to Summarize a Congressional Hearing in < 5 Minutes

Testing LLM tools on basic legislative office tasks.

A key task for legislative office interns is to attend and summarize hearings. I decided to see how well new large language models could assist with this, using the example of the September 14 House Oversight Subcommittee Cybersecurity, Information Technology, and Government Innovation hearing in the on “How are Federal Agencies Harnessing Artificial Intelligence?”.

Below are the steps I took to arrive at a basic hearing memo in under five minutes. The result is pretty good, in my opinion, especially when viewed in light of current resource restrictions in Congress that mean that most offices do not have sufficient staff/time to have someone review *every* hearing.

1. Open hearing on YouTube

2. View the transcript, click CTRL+A to highlight all, and CTRL+C to copy transcript contents

3. Paste transcript contents into Anthropic Claude (I use Claude because it has a large token window and allows you to paste in a large block of text)

with the prompt “Please generate a summary and key points from this hearing transcript”

The results were pretty good:

But I wanted details — including quotes from specific Members of Congress.

4. Copy the list of Member names from the committee’s website and paste into Claude with the prompt: “Please list each member and the key points each made.”

Again, the results were pretty good:

Here are the key points made by each member of the Subcommittee during the hearing:

Rep. Nancy Mace (Chairwoman, SC-01):

- Expressed urgency around government compliance with laws on AI use policies and workforce assessment.

- Raised concerns about potential bias, privacy risks, and job impacts from government AI systems.

- Emphasized the need for oversight, legislation, and harnessing AI responsibly.

Rep. Gerald Connolly (Ranking Member, VA-11):

- Discussed the need to curb risks of AI while allowing responsible innovation.

- Highlighted current AI use cases in government like chatbots and autonomous vehicles.

- Praised Biden administration efforts on AI transparency and called for accountability.

Rep. William Timmons (SC-04):

- Asked about characterizing AI capabilities to understand realistic military applications.

- Raised concerns about potential weaponization of AI by non-state actors.

Rep. Ro Khanna (CA-17):

- Discussed differences between human and artificial intelligence.

- Proposed legislation to leverage AI for improving government website search.

- Asked about job displacement versus augmentation from AI.

Rep. Clay Higgins (LA-03):

- Raised concerns about authoritarian regimes and criminal enterprises misusing AI.

- Discussed need to balance AI’s promise and risks.

- Questioned dulling of human instincts in law enforcement uses.

Rep. Nick Langworthy (NY-23):

- Emphasized the need for transparency in agency AI use case inventories.

- Discussed potential of regulatory sandboxes to encourage AI innovation.

-Raised concerns about generative AI enabling harms to children.

As a side note, I noticed Rep. Higgins mentioned (who is not on the committee) so I asked about that (and got the right answer!):

5. I wanted to make sure to get the details of the witnesses, so I asked, “Please list and Summarize key points made by each witness” (which it did pretty well):

6. So then I asked Claude to put it all together in a memo that would be useful for a member of Congress. Here is the prompt:

You are a legislative assistant advising a Member of Congress on issues of government technology. Write a memo explainign what happened at the hearing highlighting any issues that may require future legislation and especially noting any comments from chair and ranking member, any themes emerging that demonstrate different approaches by party and any recommendations for future action

And the result:

Dear Representative,

I’m writing to brief you on the key issues and themes that emerged during yesterday’s House oversight hearing on the federal government’s use of artificial intelligence (AI). This technology presents both opportunities and risks, making congressional oversight and potential legislation important.

The chairwoman, Rep. Nancy Mace, opened by underscoring the urgent need for the Biden administration to comply with laws requiring AI use guidance and workforce assessments — an issue that signals the need for stronger enforcement mechanisms. Meanwhile, the ranking member, Rep. Gerry Connolly, acknowledged beneficial current applications like chatbots but stressed curbing risks and allowing responsible innovation — implying a balanced regulatory approach.

Their comments point to differences in how the parties are evaluating the appropriate legislative response. Republicans focused more on risks of government overreach and criminal exploitation of AI. Democrats emphasized harnessing AI safely but also its potential for improving services. While varied perspectives were raised, all agreed on the need for transparency, accountability, and human discretion.

Key issues discussed that may require future legislation include:

• Compliance enforcement — The delays in meeting AI reporting mandates could warrant penalties or funding restrictions.
• Transparency standards — Inconsistent disclosure of agency AI uses argues for legislating rigorous public reporting criteria.
• Civil liberties risks — Concerns over authoritarian regimes misusing AI may necessitate formal restrictions on government applications posing disproportionate threats to rights.
• AI weapons risks — Fears over criminal or terrorist misuses merit re-assessing checks on military and law enforcement AI weapons uses.
• Preserving human roles — Calls made to codify oversight for law enforcement, border agents, and military to retain human discretion and instincts.

This hearing highlighted important oversight issues as AI advances rapidly. I recommend we continue monitoring agency activities and explore options for balanced, bipartisan legislation that provides transparency, protects rights and security, and upholds human accountability. Please let me know if you need any other insights as you evaluate recommendations for committee action.

Sincerely, [Your name] Legislative Assistant [Representative]’s Office

And that’s how I used Claude to summarize a hearing in under five minutes!

The memo is not perfect and certainly lacks nuance. If I were advising my public policy students, I would suggest changing it to third person tense, and including key quotes from Members and witnesses. It probably would not take too many more prompts to use Claude to get there. But for under five minutes of prompting, it’s a pretty good start!

Marci Harris is a lawyer, former Congressional staffer, founder and CEO of POPVOX, Inc. and Executive Director of POPVOX Foundation. She is an adjunct professor at the University of San Francisco and a political science lecturer at San Jose State University, where she encourages students to use LLM tools in her courses.