As the 2026 US midterms approach, one story deserves more attention than it's getting: the use of AI tools to run covert influence operations. OpenAI has disrupted more than 20 operations in which its tools were used to generate fake articles, posts, and comments aimed at manipulating public opinion — traced to actors linked to Russia, China, Iran, and Israel. Analysts expect all three of the first-named to attempt to influence the midterms.
The encouraging detail: OpenAI reports that none of the operations it caught achieved viral traction or built lasting audiences. The less encouraging one: AI-generated content still shapes discourse and hardens polarisation even when individual campaigns flop. Here's how to follow this story without either panicking or dismissing it.
Why This Matters Now
- The barrier to entry collapsed. Producing thousands of plausible, native-sounding posts across languages used to need a staffed troll farm. Now it doesn't.
- Detection is improving too. Platforms and labs are catching and publishing these operations, which is why we know about them. That's a genuine counterweight worth crediting.
- The scale question is unsettled. Whether AI content actually changes votes — versus just adding noise — is contested by serious researchers. Anyone certain in either direction is ahead of the evidence.
If you follow the AI cybersecurity story, this is its information-space sibling: same capabilities, aimed at people rather than systems.
The Best Podcasts for the Story
- Disinformation and platform-research shows — podcasts featuring the researchers who actually track these networks. They're precise about what's proven versus suspected, which matters enormously here.
- National security and foreign policy shows — for the state-actor angle: who runs these operations, what they want, and how governments respond.
- Tech policy podcasts — for the platform-and-regulation side, including what labs and social networks are obliged to detect and disclose.
How to build a feed: search "influence operations," "AI disinformation," and "election security" across Spotify and Apple; prioritise episodes with academic researchers over pundits — this is a field where methodology decides whether a claim means anything.
What to Listen For
- Reach versus existence. That an operation existed is not the same as that it worked. The credible guests separate the two carefully.
- Attribution confidence. Linking a campaign to a state is hard. Listen for hedged, evidenced attribution rather than confident naming.
- The liar's dividend. A subtle harm: once people know AI fakes exist, real evidence gets dismissed as fake. Some of the best episodes focus here.
- What actually helps. Provenance standards, platform labelling, and media literacy all get proposed — with varying evidence behind them.
Don't Just Listen — Capture It
This topic is unusually easy to absorb badly. Alarming claims stick; the caveats that make them meaningful don't. A month later you're left with a vague sense that everything online is fake — which is its own kind of misinformation.
- Paste the episode link into DriftNote for a structured summary — overview, key topics, takeaways, and quotes with timestamps.
- Skim it after listening and mark which claims were evidenced and which were speculative.
- Save it in Notion so you can track how the research actually develops through the election cycle.
A Fast Listening Plan
- Start with a researcher-led episode on how these operations are detected.
- Follow with a national security show on the state actors involved.
- Finish with a tech policy episode on platform and lab responsibilities.
Summarize each, and you'll head into the midterms informed rather than just uneasy.
Where to Go From Here
- Try the free podcast summary tool
- When AI runs the attack: AI and cybersecurity
- The best news and geopolitics podcasts for 2026
- Notion podcast notes template
The useful position on AI and disinformation is neither panic nor dismissal — it's precision. Listen to the researchers, keep the caveats, and you'll be a much harder target than most.