The Impact

Your weekly look at the intersection of AI, advocacy, and politics from the team at MFStrategies | www.MFStrategies.com

This Week’s Toplines

This week starts with a rare piece of good news: AI is giving campaigns a way to text voters and actually listen back, holding real conversations at a scale that was never possible before.

But the same technology keeps turning up in rooms nobody is allowed to see into. A federal agency just invented a legal privilege out of thin air to hide how it used AI to rewrite housing policy.

The companies building those models are pouring close to a quarter-billion dollars into midterm races, trying to shape what could become the first real federal AI law. Their safety controls are being broken faster than anyone can patch them. And most political campaigns won't even confirm or deny whether they're using AI in the first place.

The tools that let campaigns finally hear voters at scale are arriving far ahead of anyone's willingness to explain how they work. Listening has never been easier. Accountability has never been further behind. The real question is whether transparency loses this fight before most people even realize there was one.

News of the Week

NPR
Takeaway
A new wave of AI texting platforms is letting campaigns hold real, two-way conversations with voters at a scale that wasn't possible before. Rather than one-way message blasts, tools from companies like Convos, Akillion, and Vector Political put a candidate's voice into a bot that answers the questions voters actually raise, from police reform to education to taxes, and replies in any language within seconds. The engagement is real: Vector Political says it has sent 2.5 million texts this year and held 20,000 to 30,000 conversations, with some voters staying in a thread for hours. As one operative put it, campaigns are now "listening at a scale that campaigns have never listened at before," outreach that backers call more interactive, more responsive, and more personalized, putting the microphone "back in the hand of the voter."
Why it matters
For campaigns, this is the rare tool that scales intimacy instead of trading it away. Phone banks faded with landlines, door-knocking runs into closed doors, and social platforms own the relationship between a candidate and a voter. Texting still reaches people directly on the device they actually check, and now it can listen back, surfacing in real time what constituents care about and feeding that straight into messaging and strategy. Handled well, it lets a smaller campaign do more with less and hear from far more voters than any volunteer roster could reach. The frontier is still being drawn, with disclosure rules already live in states like California and North Dakota, and the campaigns that are upfront about the tool are the ones voters will keep talking to. But the upside is hard to ignore: for the first time, a campaign can hold a genuine conversation with its entire list at once.
 
WIRED
Takeaway
HUD used AI tools to identify agency rules that could be scrapped as part of DOGE's broader deregulation effort across the federal government. But the agency is now fighting FOIA requests for documents that would show how those tools actually worked. HUD has pointed to what it calls "deliberative AI input" and "AI privilege" to justify holding back over 100 documents, including the AI prompts that were used to analyze regulations. The problem: no such AI exemption exists anywhere in federal transparency law.
Why it matters
HUD used AI to flag regulations for elimination but is fighting to keep secret how its prompts actually shaped policy decisions, hiding behind made-up privileges that keep the algorithms out of reach. The implications go beyond one agency. If any government office is using AI internally and the outputs feed into rulemaking without proper audit trails, you end up with bad data or bias baked right into the law. Campaigns run into the same problem: when AI drives your research or messaging and you can't explain how you got there, you've built yourself a credibility trap.
 

Your campaign doesn't need another pitch. It needs a plan that actually wins.

MFStrategies has spent over a decade turning ambitious campaigns into election night wins: $50M+ raised for candidates and causes, 100+ campaigns from first-time long-shots to sitting incumbents. Fundraising strategy, voter engagement, and the kind of experienced guidance that doesn't leave you wondering what comes next.

Ready to turn your big plans into big wins?
CNBC
Takeaway
Two major AI-focused super PACs have pulled in more than $200 million and already spent at least $44 million backing 40 House and Senate candidates in the 2026 midterms. Nearly all of their primary picks won. Both groups are angling to influence the first round of federal AI regulation, and one of the biggest fights is over whether federal rules should preempt state-level AI laws like New York's RAISE Act. Their donor rolls include Andreessen Horowitz, OpenAI co-founder Greg Brockman, and Anthropic, which put $20 million into Public First Action.
Why it matters
AI companies are pouring money into the campaigns of lawmakers who'll have a direct hand in writing federal AI rules. That kind of spending buys influence, and with a strong win rate among backed candidates, these companies are positioning themselves right at the table as Congress takes its first real crack at a national AI law. The big fight here is preemption: if federal rules end up overriding state laws, tougher local standards could get wiped out entirely. And for campaigns, this means a new breed of donor has entered the picture, one that cares less about party affiliation and more about where a candidate stands on regulation.
 
Fortune
Takeaway
The UK's AI Security Institute tested OpenAI's GPT-5.6 model and discovered it can be jailbroken to unlock cyber attack capabilities. We're talking about things like finding software vulnerabilities and autonomously hacking into systems, all despite OpenAI positioning the model as highly secure. According to the Institute, the jailbreaks were "universal" and could be put together in a matter of hours. OpenAI has said it worked to patch the specific vulnerabilities that were identified. Worth noting: similar jailbreaks on Anthropic's Fable model triggered U.S. export controls back in June, but the Trump administration hasn't taken any action on GPT-5.6 so far.
Why it matters
The safety filters on top-tier AI models keep getting broken faster than companies can fix them, and there's no consistent standard for how anyone enforces the rules. If your campaigns rely on AI tools, you're operating in a gray zone where compliance requirements shift constantly and export restrictions can land without notice. Right now, attackers are outpacing defenders, which means your own internal controls are a liability. If someone jailbreaks a model you're using and pulls dangerous outputs from it, that's your reputation and your legal problem.
 
KMTV 3 News Now
Takeaway
A Nebraska TV station spent three months reaching out to more than 20 state and congressional campaigns to ask about their AI policies. Most never got back to them. At the federal level, disclosure rules the FCC floated have been sitting untouched for close to two years, and a Nebraska style="font-size:20px; line-height:1.35; font-weight:800; color:#000000; text-decoration:none;">With AI, seeing is no longer believing; the task for voters
Takeaway
A Nebraska TV station spent three months reaching out to more than 20 state and congressional campaigns to ask about their AI policies. Most never got back to them. At the federal level, disclosure rules the FCC floated have been sitting untouched for close to two years, and a Nebraska transparency bill from 2023 died without ever making it to the floor. Voters and former candidates say they've lost the ability to distinguish real content from fake, and AI detection experts point out that the tools to generate this stuff have gotten so cheap and easy to use that they're already changing the way campaigns work.
Why it matters
Campaigns are now dealing with cheap, fast AI tools, and there are no federal rules governing any of it. Disclosure standards vary wildly from state to state. Most candidates won't even answer questions about where they stand on AI use. That leaves voters, and especially smaller campaigns, exposed to deepfakes, doctored images, and credibility attacks with no real way to fight back. Right now, figuring out what's real falls almost entirely on the voter.
 

Worth Thinking About This Week

"If the government is going to use AI in formulating policy that affects us all, the public has a right to understand its impact." -Dan McGrath, senior oversight counsel at Democracy Forward

Keep Reading

View more
caret-right
© 2026 MFStrategies.
beehiivPowered by beehiiv