Powerful new AI systems are emerging that could dramatically accelerate biological research, for good and for ill, but mainly ill. Companies like Anthropic are now openly restricting public and independent access to their most advanced models, citing biosecurity risks. The message is clear: these tools are too dangerous for ordinary scientists, small labs, entrepreneurs, or curious citizens. But they are apparently safe enough for governments, approved institutions, and large corporations.

This is not a minor technical decision. It is another step in the institutional inversion we see across the West: elites and bureaucracies claiming exclusive control over powerful technologies while telling the rest of us to trust them with the keys.

Robert Malone has laid this out sharply. The same federal apparatus now positioning itself as the guardian against AI-enabled biothreats spent years funding, defending, and obscuring controversial gain-of-function research. Decades of grants, subcontracts, and international lab networks, including work connected to the origins of SARS-CoV-2, only came to light through relentless congressional pressure, whistleblowers, and litigation. Transparency was not offered; it was dragged out kicking and screaming.

And what has been the result of that oversight? Hearings, reports, strongly worded letters, but little fundamental reform. The same agencies continue operating with similar architectures. The same incentives for secrecy and self-protection remain. Now we are told these institutions should have privileged access to AI systems that could supercharge biological work, while independent researchers are locked out "for our safety."

This is the same pattern we see with grooming gangs in the UK, two-tier policing, speech restrictions, and migration policy: institutions that no longer see themselves as servants of the public, but as managers of a population that cannot be trusted with dangerous knowledge or uncomfortable truths.

Governments and their favoured partners have a miserable track record on biosecurity. Recent cases of biological materials being smuggled or mishandled by researchers only underline the point. If the system cannot reliably control physical pathogens and lab materials, why should anyone believe it will exercise wiser judgment over AI systems that could design or analyse them far more efficiently?

The deeper problem is structural. Modern government is not a unitary, competent actor. It is a sprawling network of agencies, contractors, universities, and international partners with misaligned incentives, conflicting agendas, and a long history of prioritising institutional self-preservation over accountability. Concentrating advanced AI capabilities in such hands does not reduce risk, it merely relocates and magnifies it.

Meanwhile, attempts at control are largely performative. Open-source and lower-cost models (such as those coming out of China) show that frontier capabilities cannot be permanently contained within a handful of Western corporations and government-approved labs. Knowledge spreads. Technology proliferates. Restricting access for Western citizens and independent researchers simply ensures that adversarial states and well-resourced actors will advance while domestic innovators are hobbled.

The biosecurity argument is not without merit: genuinely dangerous dual-use capabilities exist. But when the proposed solution is to hand monopoly control to the very institutions whose failures helped create the current climate of distrust, scepticism is the only rational response.

This fits the larger epistemological crisis: elites demand trust and deference precisely as their track record justifies the opposite. Whether it is pandemic policy, climate modelling, financial regulation, or now advanced AI, the pattern is the same: centralise power, restrict public scrutiny, and label dissent as dangerous or misinformed.

Ordinary people: farmers, small business owners, independent researchers, parents, are once again being told they are too irresponsible to handle powerful tools. The same people who just want to work, raise families, and speak plainly without being monitored or managed.

We have been here before. The instinct to control information and technology "for our own good" has a long and undistinguished history. It rarely makes us safer. It reliably makes those who hold the controls more powerful and less accountable.

The right approach is not blind rejection of safety measures, but epistemic humility on all sides. Genuine biosecurity requires transparency, distributed scrutiny, parallel research pathways, and robust oversight that actually works, not performative restrictions that concentrate capability among those least deserving of unchecked trust.

Until institutions demonstrate they have learned from past failures, gain-of-function, pandemic origins, selective enforcement, the public is right to view these new restrictions with deep suspicion. Powerful AI, like other transformative technologies, should not become another tool for elite gatekeeping in an already hostile institutional landscape.

The future of powerful AI will be decided by who gets access. The question we should be asking is not only whether the technology is dangerous, but whether the institutions claiming exclusive stewardship have earned the right to decide for the rest of us.

https://www.malone.news/p/the-ai-they-dont-want-you-to-havel