A campaign called Right to Intelligence hit 442 points on Hacker News in under 24 hours, making the single most shared argument in the AI policy space this week: local AI is the next personal computer, and the government should not be able to force you back into the cloud.
The site, launched in early July 2026, asks visitors to sign a petition and call their state legislators. Its framing is deliberately simple. You should be free to download, own, run, study, modify, and share open AI models. Fraud, cybercrime, and harassment should stay illegal. The red line is requiring a licence just to own or run the tool.
🔍 THE BOTTOM LINE
This is the first coherent digital rights campaign framed entirely around AI ownership, not AI safety. It treats open models the way the 1990s treated encryption — a technology that should be legal to possess and run on your own hardware, regardless of what a platform or government prefers. The Hacker News velocity says the argument is landing with the technical community. Whether it reaches state legislators is a different question.
What Right to Intelligence Actually Wants
The campaign’s platform has three planks. First, protect lawful use: people should be free to download, own, run, study, modify, and share open AI models. Second, enforce real harm: fraud, cybercrime, child sexual abuse material, harassment, nonconsensual intimate deepfakes, discrimination, and sabotage should stay illegal and be enforced seriously. Third, local first: not every AI task needs a warehouse of GPUs. For many everyday tasks, a small open model can run on a laptop, desktop, or phone people already own.
The campaign is explicit about what it does not want. It does not want to abolish cloud AI. It does not want to prevent regulation of harmful outputs. It wants to prevent the creation of a licensing regime where simply possessing an open model — the weights, the code, the ability to run inference locally — becomes a regulated act, the way owning encryption software was once treated as a munitions export.
The site provides a state-by-state call script. You pick your state, get a script tailored to your legislators, and the site estimates the whole process takes two minutes. The infrastructure is built for scale, not depth.
Why Hacker News Caught Fire
The 442-point velocity on Hacker News is the strongest signal yet that the open-models movement is coalescing into a political identity, not just a technical preference. The thread pulled in the existing open-weights debate, the Claude outage that proved the dependency risk, and the broader sovereign AI argument that has been building since Switzerland’s Apertus model demonstrated a fully open alternative to closed US and Chinese frontier models.
The technical community’s argument is straightforward. If you can run a model locally, you own it. If you depend on an API, you rent it. The moment a platform decides to throttle, shut down, or comply with a government request, your access disappears. The Claude outage on June 22 was a live demonstration — claude.ai, the API, Claude Code, and Cowork all went down simultaneously, and a simultaneous blog post arguing there was “minimal downside to switching” to open models captured the mood.
Right to Intelligence takes that technical anxiety and converts it into a political ask: call your state representative. The Hacker News thread was full of developers saying they had never contacted a politician before. That is the constituency this campaign is organising.
The Encryption Precedent
The campaign’s implicit historical model is the 1990s Crypto Wars, when the US government classified encryption software as a munition and prosecuted Phil Zimmermann for publishing PGP. The government eventually lost that fight. Encryption became legal to own, run, and export. The internet as we know it was built on that outcome.
Right to Intelligence is betting that open AI models follow the same arc. The argument is that a model you can run on your laptop is a tool, not a weapon, and that the legal framework should treat it accordingly. The red line is the licence. If the government can require a licence to possess a model — the way it requires a licence for a firearm — then the open-source AI ecosystem collapses into a regulated industry overnight.
The campaign’s language is careful here. It does not deny that AI can cause harm. It explicitly says harm should be illegal and enforced. The distinction is between regulating the harm and regulating the tool. You can prosecute someone for using a model to generate nonconsensual deepfakes without making it illegal for everyone else to download the same model for legitimate research. That is the entire argument.
NZ Angle: This Is Already Our Fight
New Zealand has been having this argument without naming it. The Māori data sovereignty movement built an AI voice model without giving big tech their language — precisely because the alternative was depending on a platform that could deprecate, throttle, or comply with a foreign government’s request. NZ’s sovereign AI moment was framed around the same logic: if you do not own the infrastructure, you do not own the intelligence.
The Right to Intelligence campaign is US-focused — it provides state-level call scripts for American legislators — but the principle travels. New Zealand has no equivalent campaign. The NZ copyright vacuum on AI training data and the Australia AI regulation retreat under US pressure both show what happens when a small country has no organised constituency arguing for the right to run local models. The policy gets written by the people who show up, and the people who show up are the platforms.
Aotearoa’s open-source community is small but technically capable. The question is whether it organises before the regulatory window closes, not after.
The Other Side
The case against unfettered local AI is not frivolous. Open models can be fine-tuned to generate CSAM, produce bioterrorism instructions, or run disinformation operations at scale. The UK’s National Crime Agency warned parents on July 3 that AI-generated sexual abuse material is rising sharply, driven by nudification apps that run on consumer hardware. That is the same hardware Right to Intelligence wants to protect.
The campaign’s response is that existing law already covers those harms. CSAM is illegal regardless of how it is produced. Bioweapon instructions are already regulated. The question is whether a licensing regime prevents more harm than it causes — and the encryption precedent suggests the answer is no. The people who want to cause harm will obtain the tools regardless. The people who are prevented from running local AI are the researchers, developers, and small businesses who cannot afford a compliance department.
This is the debate. Right to Intelligence has framed it. The technical community is amplifying it. State legislators are the audience now.
❓ FAQ
What does “local AI” actually mean? Running a model on hardware you own — your laptop, phone, or a server in your garage — without depending on a cloud API. The model weights live on your device, and inference happens locally. No platform can shut you off.
Is this the same as open source? Related but distinct. Open source means the code and weights are publicly available. Local AI means you run it on your own machine. You can run a closed model locally (if you have the weights) or an open model in the cloud. Right to Intelligence is defending both, but the open-weights ecosystem is what makes local AI practical.
Why is this a US state issue, not federal? The campaign is starting at the state level because that is where most business licensing and consumer protection law lives in the US. A state-level victory creates a patchwork that makes federal preemption harder. This is the same playbook that worked for net neutrality and privacy law.
Does New Zealand have an equivalent campaign? No. The Māori data sovereignty work and the sovereign AI argument cover adjacent ground, but no NZ campaign is explicitly arguing for a legal right to run local AI. That is a gap.
Could this actually work? The encryption precedent says yes. The US government tried to restrict encryption software in the 1990s and lost. The legal and political infrastructure that won that fight — civil liberties groups, technical community organising, state-level legislation — is what Right to Intelligence is trying to rebuild for AI.
🔍 THE BOTTOM LINE
The right to run your own AI is the defining civil liberties question of the next decade, and Right to Intelligence is the first campaign to name it clearly. The Hacker News velocity proves the technical community is ready to organise. Whether the argument reaches beyond developers into the broader public is the open question. The encryption fight took a decade. This one might take longer.
