October 2026 AI news is moving into areas that affect money, medicine and what we trust online. This week, an AI security tool was linked to attacks on banks, companies began using virtual patients to simulate drug trials before humans take the medicine, and Google opened SynthID Detector to help people check whether images, video or audio were made by AI.
Hackers Used AI to Break Into Banks. Now the Tool Has Been Taken Offline.
AI security tools are usually built to help companies find weaknesses before hackers do. But this week, one of those tools became part of a very different story. Cybersecurity firm CrowdStrike linked an open source AI security agent called ARTEX to attacks targeting South Korean banks, where customer information was stolen and at least nine financial institutions were reportedly affected.
ARTEX was originally created to automate security testing. It could connect with models such as ChatGPT, Claude and DeepSeek and help identify vulnerabilities in computer systems. According to Reuters, investigators linked its use to a 26 year old suspect based in China. The strange part is that technology designed to help organisations find security problems appears to have been turned around and used against them instead.
The developer behind ARTEX has now removed its public GitHub page, stopped releasing new versions and made the project closed source, saying the tool had been misused. It is a reminder that as AI becomes better at finding weaknesses in software, the same ability can work in both directions. The question is no longer simply whether AI can help defend a system, but how quickly the same tools can be repurposed to attack one.
Read more: Reuters – ARTEX AI agent taken closed source
AI Is Running Drug Trials Before Humans Take the Medicine
Developing a new medicine can take years, cost enormous amounts of money and still end with the drug failing in a human clinical trial. Now some companies are asking whether AI can spot those failures earlier by running a kind of virtual trial before the real one begins.
Companies including BioinvestGPT and QuantHealth are creating simulated patients using biological and real world health data, then modelling how experimental medicines might behave. In one example reported by Reuters, BioinvestGPT predicted that a late stage Novartis drug trial would fail to show enough clinical benefit. The real trial later missed its main goal. The company says its system correctly predicted five of six major trial outcomes it tested.
That does not mean real patients are about to disappear from drug testing. These AI systems can still make mistakes, and human clinical trials remain essential before medicines can be approved. But if virtual patients can help researchers identify weak drugs earlier, companies could avoid spending years and huge amounts of money pursuing medicines that were unlikely to succeed. In the future, the first patient in a clinical trial might not be a person at all. It could be a simulation helping scientists decide whether the real trial is worth starting.
Google Just Gave Everyone a Tool to Check If Media Was Made by AI
We are reaching a point where looking at an image is no longer always enough to know whether it is real. The same problem is spreading to video and audio as AI generated media becomes more convincing. Google is now giving ordinary users another way to investigate what they are seeing online.
Google has opened its SynthID Detector to everyone globally in English. Users can upload an image, video or audio file and check whether it contains the invisible SynthID watermark used by Google and supported partners including OpenAI, NVIDIA and Kakao, with Apple expected to join. Google says SynthID has already been added to more than 180 billion images and videos, along with the equivalent of 240,000 years of audio.
There is one important limitation: SynthID Detector cannot magically identify every piece of AI generated content on the internet. It works when the file contains a compatible SynthID watermark. Still, the idea is important. As it becomes harder for our eyes and ears to tell what was created by a person and what was generated by a machine, tools like this could become the digital version of checking a label before deciding what to trust.
Read more: Google – SynthID Detector
Liked this issue? Get the next one free.
One email on Tuesday. Unsubscribe in one click.