Americans trust AI for politics more than money, new survey finds

This juxtaposition—Americans deeply skeptical of AI touching their retirement accounts but relatively blasé about AI‑shaped political content—is a flashing red light for anyone worrying about societal alignment. On the financial side, people seem to intuit that opaque, probabilistic systems can make costly mistakes and that some domains still require human accountability. In politics, by contrast, the same caution hasn’t kicked in: people will happily consume AI‑curated or AI‑generated messaging without demanding the same verification they expect for financial advice.([journalrecord.com](https://journalrecord.com/2026/01/01/americans-ai-finance-politics-trust/))

For the race to AGI, this matters because it hints at where advanced systems will find the path of least resistance into human decision‑making. If safety debates focus narrowly on catastrophic misuse while neglecting slow‑burn shifts in political attitudes, AI labs and platforms might inadvertently optimize models for persuasive power over truthfulness. The studies cited in the piece already suggest that more "persuasive" chatbots can nudge political views in durable ways, even when they occasionally hallucinate. That dynamic amplifies the incentives for bad actors to weaponize increasingly capable models long before we reach anything like AGI, and it raises the bar for alignment work that explicitly constrains models’ ability to manipulate beliefs.