Deceptive AI Act (H.R. 8893)

Summary

Directs a government agency to create expert groups that will develop rules for labeling content made by artificial intelligence to protect people.

What problem does this solve?

It is becoming very hard to tell the difference between real content and content made by AI, which can be used to trick people. This bill creates expert groups to develop ways to label AI content, like watermarks, so everyone can easily see what is real.

What does this bill do?

Establishes AI content task forces

Requires the National Institute of Standards and Technology (NIST) to create special groups, or task forces, to help develop ways to identify content made by generative AI.

Develops standards for identifying AI content

Directs the task forces to support creating technical standards like watermarking, digital fingerprinting, and content history data for audio, video, and text made by AI.

Includes a wide range of experts

Mandates that the task forces include members from government, AI companies, social media platforms, academic groups, privacy advocates, media organizations, and labor unions.

Requires reports to NIST and Congress

Requires the task forces to give their recommendations to the Director of NIST and to send yearly reports to Congress for five years on their activities.

Considers user privacy

Instructs the task forces to create guidance for online services to handle content history data in a way that protects privacy and gives users control over their information.

Who does this affect?

Technology companies and AI developers
Social media and online content platforms
General public and internet users

What is the real world impact?

•

Increases transparency for online content

Aims to help people know when they are looking at, listening to, or reading something made by AI. This can stop the spread of false information and protect people from being tricked by fake content, often called 'deepfakes'.

•

Addresses privacy concerns with content tracking

Requires the new standards to protect user privacy. The groups must create guides on how companies can show where content came from without sharing private information, and give users options to limit what data is shared.

•

Faces challenges with enforcement

While the goal is to label AI content, people who want to cause harm will likely try to remove watermarks or get around these rules. The task forces are asked to think about these issues, but stopping determined individuals will be difficult.

When does this start?

This bill sets several deadlines for actions that must be taken after it becomes law.

Establishment of task forces

The National Institute of Standards and Technology must create the task forces within 90 days of the bill becoming law.

Recommendations to NIST Director

The task forces must submit their recommendations for standards and guidelines to the NIST Director within 270 days of being created.

First report to Congress

The task forces must send their first report on their activities to Congress within one year of being created.

Annual reports to Congress

After the first report, the task forces must continue to send a report to Congress every year for five years.