Will AI safety's media surge convert into real political power?

Jasmine Sun argues that although AI safety "went mainstream" in September 2026 — she cites Jacob Coxon on Fox, Daniel Kokotajlo on Rogan, and Nate Soares on Tucker, plus a Coxon tweet reposted by Sheryl Crow and Maggie Rogers that sits at "a staggering 175 million views" — that surge in visibility is unlikely to translate into either grassroots mobilization or electoral weight. She reports that "three-quarters of Americans are more concerned than excited about AI" and that "77 percent of Americans now want to slow or stop AI development until its safety can be evaluated," yet says she is "tentatively skeptical" that these numbers convert into organized politics.

Her reasoning runs through several empirical observations from her own reporting and from polling. First, the data-center backlash she spent the summer covering is "an 80/20 issue" but motivated by "material quality-of-life concerns" rather than ideological opposition to AI — most opponents she interviewed "did not bring up AI, AI products, or AI companies" unprompted, so that energy doesn't automatically transfer to broader anti-AI activism. Second, unlike Uber's Prop 22 campaign or Airbnb's host lobbying, AI companies cannot easily mobilize users into advocates because usage and sentiment diverge: "even 56% of AI users believe AI is 'a bad thing' for society," and Gen Z excitement "dropped by 14 points to 22% in a year." Third, issue-salience polling still ranks AI low as a voting issue — only 8% named it a top-two issue in an Echelon Insights poll, and Blue Rose data put it at "#22 of #39" despite AI rising in salience faster than anything else. She adds a fourth dynamic: partisan polarization (Trump's "HOAX-posting," widening partisan gaps in concern versus excitement) threatens the "Secret Congress" condition that writers like Yglesias and Anton Leicht have argued lets technocratic safety regulation pass quietly, out of the partisan spotlight.

This sits downstream of the governance debates the nearest notes track in the abstract. Should AI legislation wait for demonstrated risks to emerge? makes the case for waiting on legislation until risk is demonstrated; Sun's reporting describes the political terrain such legislation would actually have to clear once that risk arrives, and finds it largely indifferent or incoherent rather than ready to act. The "Hugging Face hack" she names as one mainstreaming trigger is the same incident grounding Can AI alignment evaluations reliably catch misaligned behavior?; her piece traces that incident's effect in media and politics (down to Kokotajlo's own Rogan appearance) rather than in the lab. It also complicates How do we stop AI systems once they are already deployed?: a "Law of Stop" presupposes political capacity to act, and Sun's account of low issue salience and user ambivalence suggests that capacity isn't yet there even as elite and media attention rises.

The excerpt is a single journalist's synthesis of summer reporting and secondhand polling, not a study of causal mechanism — it does not establish why usage and sentiment diverge, nor test whether elite and media salience (which Sun expects to "keep rising" into the 2028 primaries) will eventually convert into votes even though it hasn't yet. The cited figures come from different pollsters (Echelon Insights, Blue Rose) rather than one comparable survey, so the rankings aren't directly commensurable. The implication she draws, cautiously, is that AI safety's media breakthrough is necessary but not sufficient for policy change, and that rising partisan polarization could foreclose the quiet, bipartisan "Secret Congress" path to regulation before public mobilization ever catches up.