Why AI systems may never be secure, and what to do about it
[
](https://www.economist.com/)
[
](https://www.economist.com/)
undefined undefined
Subscribe to The Economist
Unlock unlimited access to all our award-winning journalism, subscriber-only podcasts and newsletters
Subscribe to The Economist
Unlock unlimited access to all our award-winning journalism, subscriber-only podcasts and newsletters
Science & technology | Computer security
A “lethal trifecta” of conditions opens them to abuse
Sep 22nd 2025|7 min read
THE PROMISE at the heart of the artificial-intelligence (AI) boom is that programming a computer is no longer an arcane skill: a chatbot or large language model (LLM) can be instructed in simple English sentences. But that promise is also the root of a systemic weakness.
Already have an account?Log in
Continue with a free trial
Get full access to our independent journalism for free
Or create a free account to unlock just this article
Explore more
More from Science & technology
How good is infant formula?
Perhaps less than its marketers say
How plants keep tabs on the competition
They grow faster when their rivals are doing the same
Ancient DNA is rewriting the history of plague
Dense cities do not seem to have been necessary for outbreaks of the disease
The chocolate industry is built on the labour of bloodsucking midges
Researchers have now confirmed how cacao plants are pollinated
The coming El Niño could be the strongest ever recorded
That is bad news for a warming world
A new drug targets one of cancer’s master switches
The pancreatic-cancer drug could be the first of an entirely new class of treatments
readers loved this