Humans in the loop

Californians say how AI should be built, used, and governed

We asked people some of the hardest questions facing the state:

How is AI reshaping work and the economy? What should government do about it?

A public forum on AI

In May 2026, we opened a public forum on the Impact of AI on work to all Californians. In the months to follow, thousands of people across the state gave us their thoughts by taking our survey. Some survey respondents later joined facilitated group discussions held on Zoom. They shared their own experiences, and listened to their peers. Then they refined gut reactions into concrete policy ideas.

Along the way, participants didn’t always agree. They debated how far to regulate, how proposals would be funded, and how the rules would be enforced.

One theme came up consistently: We need humans in the AI policy loop. Not only for how people use AI, but for how companies build it and laws govern it.

Ideas we heard

First, from the survey

Our survey asked people what they’d like government to do about the impact of AI. We got a wide range of ideas for policies.

Then from live discussions

The AI policy ideas below came from the 2-hour live discussions over Zoom. Participants refined their policy ideas and voted on what was most important.

We grouped what people said there into 26 distinct ideas. Then we ranked them according to:

How often an idea came up across discussions (frequency), and
How much support it drew when it did (how often it was a popular idea).

Top 3 policy ideas

These themes came up the most often and got the strongest support.

Examples:
- Strengthen criminal liability for harm caused by AI
- Direct more resources for real enforcement
Examples:
- Set limits on data center noise, pollution, and use of water and electricity
- Study the impact of data centers on the environment
Examples:
- Ensure a human makes the final judgment on decisions that impact lives
- Ban automated AI in the hiring process

From one discussion:

“With social media, we didn’t hold companies accountable, and it tore down our democracy. I think that the era of saying that companies aren't responsible for how people use tools like this should end. I bet they would change policy and invest in certain tools to manage better if we ended that lack of accountability.”

Arts, entertainment and media worker from the Bay Area
“To your point, Section 230 of the CDA made it so that a media company would be actively punished if they tried to police user content, rather than making it something that would align their values to the public good. So, if we're looking at incentive structures holistically: where can we structure the rules so that following the rules works towards the public good and to the bottom line, rather than creating tension between them?”

Information technology worker from the Central Coast
“One thing could be who's accountable for the outputs. If there's a human evaluation, then the human is accountable, because a human should be the final decision maker. But the way AI is being used, if human evaluation is not involved, then who's accountable for the mistakes that AI is making?”

Finance worker from the Bay Area

“I think we should protect our natural resources, and a clean environment. Because I know people in Texas where data farms or data centers have moved in, and the water has been polluted, and electric bills go way up.”

Arts/entertainment/media worker from Southern California
“My biggest concern is the amount of water and electricity AI is using. You hear of the data center in Georgia that illegally tapped into some city’s or county's water supply and took millions of gallons from it without anybody realizing.”

Government worker from LA
“When I think about the state as a physical place, we could set environmental regulations, we could set data privacy requirements for states that operate here. I get really excited about what floor we are setting for environmental protection when it comes to energy use, space use, etc.”

Government worker from the Bay Area

More policy ideas

Strong support

Came up often and with strong support

7 policy ideas

Examples:
- Set rules for gathering, handling, securing, and deleting
- Enforce transparency and privacy protections
Examples:
- Urge companies to align with human standards
- Set guardrails for AI overuse and addiction
Examples:
- Add new fees for firing workers for AI
- Create incentives to keep or increase jobs
Examples:
- Give affected communities their say
- Create citizen peer groups
Examples:
- Urge them to retrain workers instead of laying them off
- Shift workers to underserved areas
Examples:
- Mark AI content with common words or watermarks
- Penalize companies when they don’t disclose AI use
Examples:
- Close loopholes that let them avoid liability
- Change tax laws so stock can’t be used to avoid taxation

Moderate support

Came up often, or with strong support, but not both

7 policy ideas

Examples:
- Enforce copyrights, defend IP
- Companies pay for training data into creator fund
Examples:
- Make public service ads on benefits and risks
- Guide residents about risks posed by AI, to privacy, children, etc.
Examples:
- Let students work in sandboxes to scale positive effects
- Data literacy taught in K-12 schools
Examples:
- Require consent, make opt-out the default
- Let people opt out of AI using their data
Examples:
- Fund studies on how AI use affects critical thinking
- Find and publish more evidence on the harm and risks
Examples:
- Make consent text simple
- Limit collection of data in public settings, so AI can’t use it
Examples:
- Universal basic income
- More unemployment pay if you lost your job to AI

Some support

Came up less often and with less support

9 policy ideas

Examples:
- Offer free AI training for state employees and expanded out to the public
- Research programs to figure out how best to retrain workers
Examples:
- Use AI to crack down on fraud and ensure benefits go to the right recipients
- Government should license new AI models
Examples:
- Preserve role of human decision makers
- Establish guidelines for human evaluation
Examples:
- Require easy-to-use parental controls on tools children can access.
- Oversee how technology companies expose schoolkids to AI tech.
Examples:
- Fund third spaces and arts/sports programs to promote community health
- Fully fund high-quality knowledge and information services like education, libraries, mental healthcare
Examples:
- Make sure students disclose using AI in schoolwork
- Check and limit AI use by students
Examples:
- Require employers to include employees in decisions to introduce AI tools
- Restrict companies from firing people who don’t want to use AI tools
Examples:
- Transparent reporting on AI’s effect on jobs lost
- Continuous reporting on AI’s effect on the economy
Examples:
- Align AI use to places where it’s socially beneficial
- Have AI do jobs people can’t do or don’t want to do

How to read this ranking

The ranking shows how an idea fared across discussions. It's not a verdict on its merit.

These are the themes participants talked about. They are separate from what current policy or legislation may already do. Some may align with items already moving through the policy landscape.

Check our ranking methods

Who participated

Survey

Broad listening

People responded to our open survey on AI

Scatter plot of dots, one per survey respondent, colored by region.

Discussions

Inviting a balanced audience

We selected a subset of respondents that reflects California

Scatter plot of dots, one per invited respondent, colored by region.

Discussions

Joining the conversation

These Californians RSVP’d “yes” and joined a discussion

Scatter plot of dots, one per discussion attendee, colored by region.

Discussions

The discussion groups

Here are the demographics of each discussion

Ring of 14 small clusters of dots, one per conversation, colored by region.

What happens next

This report is not the end of the story. The ideas people shared will join a fast-moving policy conversation already underway.

California has been a leader in AI oversight for some time. In 2025, California enacted SB 53, the Transparency in Frontier Artificial Intelligence Act. It sets requirements for transparency and independent oversight of advanced AI models.

In May 2026, the Governor signed an Executive Order to help prepare workers and businesses for changes AI may bring. The order directs state agencies to use findings from this Engaged California engagement in that work, where they can. Californians can hold the state to that promise.

As Engaged California summarized what people said, the Newsom administration acted further on AI. Some of the actions by the administration touched on themes that Californians raised in these discussions, such as:

The environmental footprint of data centers
Transparency about AI-generated content
Independent oversight of frontier models
Protections for workers and children

The actions included Governor Newsom signing the following bills and an Executive Order in September 2026:

One of the nation's broadest data center laws. This gives communities more say over use of their water, electricity, and land.
Senate Bill 1050, which requires disclosure when an ad uses AI-generated performers
An Executive Order to speed up the state’s AI oversight efforts. This fast-tracks new AI auditing laws and advances a "kill switch" for frontier AI models.
A set of first-in-the-nation AI safeguards
Child safety laws for chatbots and social media

We're not done with our work either. Californians raised many ideas that will inform government action in the years ahead. And as the AI landscape keeps evolving, we will keep listening.

Humans in the loop

A public forum on AI