Many organizations are already struggling to cope with the sheer volume of cyber attacks targeting them. This problem is likely to worsen significantly as an increasing number of criminals try their luck, and as the capabilities of publicly available generative AI systems continue to improve. Earlier this month, AI company Anthropic claimed that Mythos, a model it developed and is now testing, has found thousands of critical vulnerabilities, some of which are in every major operating system and web browser. Bushra They say They’ve all been patched, but it’s delaying the release of models resulting from the new capabilities and forming a consortium of tech companies called Project Glasswing that it says will try to make the capabilities work for defense purposes in the meantime.
Right now, cybersecurity researchers are optimistic that slow attacks can be thwarted by basic defenses, highlighting how important it is to keep software updates on top of things and stick to network security protocols. How well we will be positioned to fend off more sophisticated attacks in the future is less clear.
The good news is that AI is being used for defense as well. Every day, Microsoft — just one of many businesses monitoring such threats — processes more than 100 trillion signals through its AI systems as potentially malicious or suspicious. The company says that between April 2024 and April 2025, it managed to block $4 billion worth of scams and fraudulent transactions, many of which may have been aided by AI content. The very technology that makes such attacks possible may also be our best bet to keep us safe in the years to come.