In February this year, Anthropic, an American Artificial Intelligence (AI) company, refused to let the U.S. government use its powerful AI tools for mass surveillance and murder. Also in February this year, Anthropic was declared a threat to national security. And get this, by the government of its own country. Blindsided by the rapid progress that AI labs are making, the U.S. government is rushing to incorporate AI into its workflow — even before policies and restrictions regarding its use are ready — and is ready to attack companies that oppose.

Yes, modern AI is well-equipped to perform mass surveillance, and would probably be the best candidate for the job; therein lies the danger. Much of the power AI provides stems from its speed and ability to perform large-scale tasks, effectively reducing much of the friction that comes with doing so by hand. AI allows the average human to do what only trained professionals could do before at much faster speeds. The power that those who wield the latest AI models gain is at uncharted levels, allowing people to do whatever their heart desires at scale (including, cough, cough, mass surveillance or autonomous murderbots).

This is made infinitely worse because AI policy is lagging: no one knows what to do because we have simply never seen anything quite like it before. Current U.S. government policies just aren’t well-equipped to protect citizens from bad-faith government work, especially when it comes to cutting-edge technology. Before we enforce strict rules regarding the use of AI, it’s all up to humanity.

And humanity is… quite a force to be reckoned with.

They say that power corrupts, and with unchecked power, even democratic governments can easily implement mass surveillance or censorship policies. These policies give governments significant control over our social lives, allowing them to better pursue their agenda. However, it need not be said that it’s obviously terrible for the citizens. In particular, the Trump administration has empowered surveillance companies such as Palantir to collect and aggregate the personal data of Americans into governmental databases. This gets worse when you realize that the LLM models are being used to identify so-called “suspicious activity,” such as criticisms of the Trump administration. You can imagine how this is being used. According to 404 Media, Palantir’s app, ELITE, “populates a map with potential deportation targets,” allowing Immigration and Customs Enforcement (ICE) to “find locations where lots of people it might detain could be based.”

Just last year, Taiwan’s Ministry of National Defense established an Artificial Intelligence Office to explore AI-related military applications. Déjà vu?

Furthermore, according to Taiwan News, the Taiwanese government is seeking to aggressively incorporate AI into government work to help automate and ease workloads, even as clear, objective guardrails are falling further and further behind.

No government, even Taiwan, which ranks 26th in happiness according to the 2026 World Happiness Report, is safe from this. No matter how strong their moral compasses are, there comes a point when their desire for control and mass automation with AI outweighs the public’s need for privacy and freedom. Without guardrails in place, the public is extremely vulnerable to those with the most power to make or break the entire country.

So, AI, certainly the landmark technology of the decade, may be pushing the frontier forward, but it’s putting everyone in an extremely vulnerable position. While the use of AI to automate repetitive tasks is exactly how it’s supposed to be used, it is reckless to assume that the government can control itself when given the shiniest, newest toy in the playground and handed little to no rules — it may speed up work, but at a cost that the public will eventually have to pay.

We need to pull the emergency brakes: the incorporation of AI into governments needs to slow drastically until we have proper legislation governing its use in place. Appropriately, as well, budgets for human resources need to stop getting cut, and layoffs need to slow down. Sure, governments won’t get the productivity boost, but if the question is public safety or government efficiency, I think the answer is easy.

Remember DOGE?