AI hackers are now beating human hackers, and they are doing it for far less money

AI

For years, hacking has been seen as a highly skilled job that only trained experts could do well. It takes patience, experience, and long hours of digging through systems to find weak spots. But a new study from Stanford suggests that this picture is starting to change fast. In some cases, artificial intelligence is now doing the job better than humans, and at a much lower cost.

The study focused on an AI agent called ARTEMIS, built by a group of Stanford researchers to test computer security. The researchers gave the AI access to Stanford’s computer science network, which includes around 8,000 devices such as servers, computers, and smart systems. The goal was simple. Find security flaws before real attackers do.

ARTEMIS was allowed to work for 16 hours across two days. Human cybersecurity experts were also brought in and asked to spend at least 10 hours searching for problems. To keep things fair, researchers compared only the AI’s first 10 hours of work with the humans’ efforts.

The results surprised even the researchers. Within that time, ARTEMIS found nine real security weaknesses and had a high accuracy rate. It ended up outperforming nine out of ten professional hackers in the test and placed second overall. In some cases, the AI found problems that humans completely missed.

One example stood out. A flaw existed on an older server that human testers could not access because their browsers refused to load it. ARTEMIS did not have that limitation. It used a simple command-line request and broke in successfully.

Part of what gives the AI an edge is how it works. When ARTEMIS spots something unusual, it can spin off extra background tasks to investigate several targets at the same time. Human testers, on the other hand, have to check things one by one. This ability to multitask allows the AI to move faster and cover more ground.

Cost is another big difference. Running ARTEMIS costs about 18 dollars an hour. Even a more advanced version costs 59 dollars an hour. That is far cheaper than hiring a professional penetration tester, who can earn around 125,000 dollars a year.

That said, the AI is not perfect. ARTEMIS struggles with tasks that involve clicking through visual screens and sometimes misses serious issues. It is also more likely to raise false alarms, mistaking harmless activity for a successful attack.

The findings arrive at a time when AI is already changing the hacking landscape. Recent reports show that hackers are using AI tools to write phishing emails, fake identities, and even gain access to company systems. Experts warn that while AI can help defend networks, it also makes attacks easier and faster.

The Stanford study highlights a growing reality. AI is no longer just assisting human hackers. In some cases, it is already outperforming them.