Stephen Miller's NSPM-7: National Security Tool or Criminalizing Dissent?
At a State Department summit on the "resurgence of political terrorism," White House adviser Stephen Miller described NSPM-7 as a directive empowering law enforcement and intelligence agencies to "disrupt, identify, defund, debank, arrest and prosecute" what he termed left-wing "political terrorists" [4][5]. The framework treats organized left-wing violence as a national security threat, opening the door to tools historically reserved for foreign terrorist organizations — designations, financial disruption, and coordinated federal enforcement.
Supporters frame this as an overdue recognition that political violence isn't confined to one side of the aisle, and that existing counterterrorism tools should apply wherever the threat is real. Critics, including Rep. Ro Khanna, warn this echoes the post-9/11 expansion of security powers that swept up legitimate activism alongside genuine threats — with the added danger that "terrorism" becomes an elastic label applied selectively to political opponents rather than violent actors specifically.
The disagreement ultimately hinges on trust: do you believe the apparatus will be applied narrowly and fairly, or do you expect mission creep in a hyperpolarized environment where yesterday's protester becomes tomorrow's "terrorist"? Both sides claim to be defending democracy — one from violence, the other from state overreach — which is precisely why the debate resists easy resolution.
Can AI Deliberation Tools Accidentally Flatten Democratic Debate?
A 2026 arXiv study on "Argument Collapse" finds that large language models used in deliberation tools tend to compress diverse, long-form arguments into a narrower band of mainstream positions — potentially sanding down minority viewpoints and unconventional framings in the process [6]. This matters because AI-mediated deliberation is increasingly pitched as a scalable solution to messy, unproductive public debate.
Related research paints a more optimistic picture: human-AI deliberation frameworks and DeepMind's AI mediator research show these tools can help people find common ground, surface areas of consensus, and even generate summary statements that explicitly incorporate dissenting views rather than erasing them [7][8]. Proponents argue this represents a genuine upgrade — better decision-making, less shouting past each other, and structured paths toward resolution in disputes that would otherwise stall.
The tension is a design question as much as a philosophical one: can AI systems be built to preserve viewpoint diversity while still helping groups converge on workable decisions, or does optimizing for "consensus" inevitably nudge outputs toward safe, conventional framing? Critics worry the latter is a structural feature, not a bug, of how these models are trained.
The Bigger Picture
Today's stories share a common thread: institutions and technologies meant to foster open discourse are themselves becoming sites of contested trust. Whether it's a free speech organization's funding sources, a government's definition of "political terrorism," or an AI's tendency to smooth over disagreement, the underlying question is the same — who gets to decide what counts as legitimate speech, dissent, or threat, and can that process withstand scrutiny from people who don't share the decider's priors?
Each case also illustrates how quickly good-faith concerns can curdle into accusations of bad faith. Toby Young sees media bias; his critics see undisclosed influence. Miller's supporters see necessary security policy; his critics see criminalized dissent. AI researchers see a tool for better deliberation; skeptics see a homogenizing force on public discourse. None of these positions are unreasonable on their face — which is exactly the kind of complexity that structured disagreement, rather than tribal shorthand, is built to handle.
Key takeaway: The hardest disagreements today aren't about facts alone, but about who controls the frameworks — funding, security law, or AI mediation — that shape which arguments get heard in the first place.
Sources
- https://www.theguardian.com/us-news/2026/jul/17/trump-administration-to-grant-12m-to-groups-founded-by-uk-conservatives-jacob-rees-mogg-and-toby-young
- https://freespeechunion.org/
- https://democracyforsale.substack.com/p/us-antiabortion-cash-floods-uk-free-speech-union-adf
- https://www.facebook.com/democracynow/posts/secretary-of-state-marco-rubio-and-top-white-house-adviser-stephen-miller-are-pu/1462730302328945/
- https://www.facebook.com/statedept/posts/secretary-rubio-under-president-trump-for-the-first-time-the-united-states-is-bu/1477123141113613/
- https://arxiv.org/abs/2606.01736
- https://arxiv.org/abs/2403.16812
- https://deepmind.google/research/publications/65220/