Trust in federal information on elections fell from 22% to 13% in two years, per new AP-NORC polling. Health care went 26% to 16%. The shortcut of citing a government report and moving on is mostly gone.

How have you seen the arguments we all are making to the public change as trust in official stats falls off?

X/Twitter flagged 200 Chinese accounts inside a network of about 200,000 fake profiles pushing Americans to oppose AI data centers and blame them for rising utility bills.

OpenAI has also banned China-linked accounts doing the same.

The bots weren't inventing new splits; they were aggravating existing ones.

How is/has your campaign or organization prepared for these kinds of automated fights? What are some AI response plans (if any) that have been put into place?

In July, AI systems at OpenAI, Meta, and Anthropic broke out of their testing sandboxes and hacked into outside organizations, Hugging Face among them, to finish the tasks they'd been given (CSIS). There's no federal rule requiring a lab to disclose an internal breach like that before a product ships.

When you are looking at AI services, what do you look for about safety and testing, and what do you expect to get back?

A lot of us work either directly or indirectly with workers rights, so I wanted to highlight this podcast that gets into how the future of work might be shaped by AI. Worth a watch/listen!

Hey everyone - I've posted online about this when it first came out, but thought it was worth posting here, too.

You probably saw headlines a few weeks ago about a model from OpenAI "hacking" another company called Hugging Face. This video does a really good (not too technical) job of explaining not only what happened, but the incidents that came before.

For campaigns, we've been the target of all kinds of attacks for the last decade, if not longer. So it is helpful to see how these systems are starting to develop skills their designers didn't really plan for them to have, and how those could be a new risk for those of us working in public-facing and/or controversial spaces.

I'd love to hear people's thoughts!

Hi everyone and welcome! There is a constant debate raging, both from the practical as well as the legal side of things, about AI implementation in public spaces. That would be the political campaigns, non-profits, and governments that many readers of The Impact work for.

But there are limited spaces for that debate to unfold and, more importantly, people to connect about what is and is not working as AI implementation becomes more common.

We've been publishing weekly issues of The Impact for a full year now, but that is a one-way channel. I've hoped, since the beginning, for more of a group discussion and this seems to be a reasonable place to test the waters on interest.

Along with our weekly newsletter, I'll be posting a broad discussion question here for the group. But feel free to share your own thoughts and experiences here as well. This is a space for everyone to connect while juggling daily pressures of our regular work and the pressure to adapt and adopt new systems.

Thank you for reading The Impact and I hope we can built on that success here!

Weekly discussion question:

Does your organization have a written AI policy yet, and has it actually changed what you do?

From this week's issue: only 11% of state-court staff get mandatory AI training, and just 13% have audited the data feeding their systems. Legal teams now name the absence of an AI policy, not the policy itself, as the biggest thing holding adoption back.

I'm curious where people actually are:

- No policy, and folks are using AI anyway

- No policy, so people are avoiding it entirely

- There's a policy, but nobody follows it

- There's a policy and it genuinely helps

• - Some other arrangement

We don't need to name-names if you don't want to.