# AI didn't invent a new attack. It made attackers more productive

Author: Norbert Biedrzycki
Published: 2026-02-26
Language: en
URL: https://norbertbiedrzycki.pl/en/ai-made-attackers-more-productive/
Summary: I think we may have just seen a preview of the future of cybersecurity, and it's not about AI "hacking systems." A hacker recently breached several Mexican…

I think we may have just seen a preview of the future of cybersecurity, and it's not about AI "hacking systems."

A hacker recently breached several Mexican government institutions and stole around 150GB of sensitive data, including taxpayer and voter-related information. During the operation, he used an AI model (Claude) as technical support.

What's interesting is not the breach itself, since governments get hacked every year. It's the economics of the attack.

The model initially refused to help; the safeguards were working. But the attacker kept trying different prompts, reframed the context as "legitimate security testing," and eventually obtained responses that helped with vulnerability analysis and automation. The campaign lasted about a month.

Could this have happened without AI? Of course. Was it likely faster and cheaper with AI? Absolutely.

And that's the real shift. AI is not creating entirely new attack vectors. It is dramatically increasing attacker productivity. Lower cost → lower barrier to entry → more actors → more attacks.

We've seen this productivity curve already in software development. Now it's reaching cybersecurity, on both sides.

There's also a bigger geopolitical layer. The U.S. government is reportedly pushing AI companies to allow broader use of frontier models in defense projects, while some companies are concerned about applications in autonomous weapons or mass surveillance.

Which raises an uncomfortable question: if advanced AI becomes critical national infrastructure, who ultimately controls its capabilities?

My prediction: incidents like this will stop being headlines. They'll become routine. Unfortunately.
