How to mitigate the expected increase in vulnerability alerts, software updates, and breaches. In recent months, the cybersecurity capabilities of Large Language Models have become increasingly powerful. Recent research shows that today’s frontier models (Claude Mythos and OpenAI GPT 5.5) with proper tooling have cybersecurity capabilities similar to trained cybersecurity practitioners and are doubling their ability every four months. Access to these frontier models has been restricted, but independent researchers have found that open weight models available to anyone can achieve comparable results. For years, we have known that software and firmware usually have many bugs, and some of those bugs may induce security vulnerabilities. AI tools can now be used to quickly find and exploit those vulnerabilities. The good news is that this capability can also be used by defenders. Companies that make software and firmware are increasingly using AI to speed the process of finding and fixing vulnerabilities in their own code. Within a few years, a much higher level of security with more secure software and firmware can be reached. However, a few years are expected to pass with elevated numbers of vulnerability alerts, software updates, and breaches. Fortunately, there are many opportunities to mitigate this risk. First, consumers and device makers should use the Matter standard. This Smart Home interoperability standard was designed with security as a top priority. Zero Trust Architecture (ZTA) principles are built into Matter to protect IoT devices against infection and minimize damage if a device becomes compromised. As an example, each Matter device gets only the minimal set of privileges that it needs. Because Matter implements the Principle of Least Privilege (PoLP), an infected light-bulb can’t cause much harm because it has low privileges. Second, device makers should protect important secrets, such as encryption keys, using chips that have certified security.