Alex Gladstein, Chief Strategy Officer at the Human Rights Foundation, says the AI debate has been hijacked. The real danger isn't a machine wiping out humanity, he argues. It's AI concentrated in the hands of a few governments and corporations — and the dictators who'd love to use it. He calls the 'AI will kill us all' line largely propaganda, a distraction from the centralization problem that's already here.
The threat model Gladstein actually worries about
Gladstein's warning isn't about a rogue superintelligence. It's about who controls the models. When AI sits in a handful of data centers owned by a handful of companies — or worse, by authoritarian states — the people with the least power lose the most. He frames centralized AI as a tool for surveillance, censorship, and control. That's the risk he says gets buried under sci-fi doom talk.
His answer: local AI. Running models on personal laptops and phones, not just through corporate APIs. For Gladstein, that's not a technical preference. It's a human rights position. If activists in a country with a repressive government can run AI locally, they don't need permission from a foreign company or a local ISP to use it.
What HRF is building and teaching
The Human Rights Foundation isn't just talking about this. It trains human rights activists to use AI, and it runs hackathons with Bitcoin Park aimed at building tools activists can actually deploy. The idea is to get working technology into the hands of people who need it, not to publish white papers.
Gladstein describes a 'Freedom Stack' — encrypted inference, open weights, and local AI. Each piece matters. Encrypted inference means nobody can see what you're asking the model. Open weights mean the model can't be quietly lobotomized or shut off by a vendor. Local AI means it runs on hardware you control. Stack them together and you've got something that doesn't depend on a corporation's goodwill or a government's permission.
Freedom money, freedom tech, and the places already using it
The same logic runs through Gladstein's work on money. He's been blunt about Central Bank Digital Currencies, describing them as a tool of repression — programmable money that can be frozen, expired, or redirected by the issuer. That's not a hypothetical for him. It's the design.
Meanwhile, he points to places where alternatives are already live. In Kenya, people are living on Bitcoin using Tando and M-Pesa. In Venezuela, after Maduro, Bitcoin has functioned as a lifeline — a way to hold and move value when the local currency and banking system fail. Gladstein also discusses building freedom tech with AI agents, tying the AI and money threads together: the same principles of local control, open access, and no gatekeepers.
The phrase he keeps returning to is 'Freedom Money.' It's the monetary counterpart to the Freedom Stack — assets and tools that don't require a central authority's blessing to work.
Where this leaves the AI safety fight
Gladstein's position puts him at odds with the loudest voices in the AI safety conversation. He's not denying that AI can cause harm. He's arguing that the harm is already happening through centralization, and that apocalyptic framing gives governments a convenient excuse to regulate and restrict. That's a fight worth watching, because the policy choices being made now — on model weights, on inference, on who gets to run what — will decide whether local AI stays possible or becomes a niche hobby for the technically inclined.
HRF's hackathons and training programs are the concrete next step. No date has been announced for the next one, but the pattern is set: build tools activists can run themselves, teach them how to use them, and keep the stack open.
