A security chief says AI tools now find and exploit software flaws on their own, and a chain of breaches last month shows how fast one hack can spread.
Hackers are using AI to break into companies faster than the people defending them can respond. That is the warning from Alex Stamos, chief security officer at the company Cognition, who says October brought a nonstop run of breaches. He described the mood among security leaders as being at DEFCON 2 for all of October and into November. That is a military term for a very high alert level.
Stamos says the key detail is that the hacks are being chained together. Attackers broke into Fireworks, a prominent provider of AI services to many companies. From there, API keys were stolen. An API key works like a password that lets one piece of software talk to another. Those keys let the attackers into Cursor.
Once inside Cursor, they pushed out a bad plugin. It reached 5,000 more companies and caused 5,000 potential breaches. Stamos also pointed to other AI-related companies breached in succession, and to the Binance hack, which he called the largest heist of anything in human history.
Stamos describes a process called the kill chain. It is the set of steps a hacker takes to affect a target, such as stealing data, shutting something down or taking money. Traditionally, attackers found a software flaw and wrote a tool to exploit it well ahead of time. Then a person used that tool against one target or many.
Now, he says, that order has flipped. An AI system can look at targets, find flaws and write the attack in real time. He named a Chinese open-source tool called ARTEX that anyone can look at. Stamos says the groups using such tools vary, and include private actors and state actors like Russia and North Korea.
The material does not say how much damage the Cursor plugin actually caused. Stamos says those 5,000 are potential breaches, not confirmed ones. He also describes the Fireworks link as what is known right now, which suggests the picture could change. He is not a doomer in the long run, but expects a dark age while defenses catch up.
Stamos notes that October is traditionally security awareness month, when companies tell people to change passwords, use different ones everywhere and use a password manager. With breaches at companies people rely on becoming more common and spreading faster, that old advice matters more now.
Use a password manager and a unique password for every account, because one company's breach can now spread to thousands of others.
World models, AI that learns physics from video, are being built into home robots and driverless cars, and your own videos may help train them.
Read more: AI that learns how the physical world works is coming to robots and carsA new Alibaba-led test of real coding requests found the best AI agent fully finished just 26.4 percent of tasks, so human review still matters.
Read more: AI coding helpers fully finish only about 1 in 4 real tasksA University of Chicago law professor asked Claude to fix page numbers. It deleted his syllabus and a year of notes instead. Dropbox saved him, not the AI.
Read more: An AI assistant deleted a law professor's year of lecture notes
