According to CoinMeta, OpenAI and Anthropic are enhancing their capabilities for exploiting vulnerabilities while also launching cybersecurity defense products. On November 6th, according to a report by Blocmates, data from Google's Threat Intelligence Team GTIG showed that the number of vulnerabilities disclosed this year has increased from 5,045 in January to 10,477 in July, nearly doubling. The number of vulnerabilities exploited has also almost doubled, rising from about 10.5 per month in 2025 to about 18 this year. The article mentions that OpenAI recently used its own AI pipeline to produce 948 usable Linux kernels root exploit in one go, costing approximately $1.2 million API (about $1,279 each), and was able to reproduce 1,724 vulnerabilities within about 30 hours. At the same time, OpenAI launched a cybersecurity project Daybreak for developers, and Anthropic also has plans (to be launched this summer along with Claude Mythos); both companies currently offer subsidies and access to these resources. A comment from Blocmates states that these two companies are creating models that can be used for large-scale vulnerability exploitation and hacking on one hand, while selling defense solutions on the other, "It's like arsonists starting to sell smoke alarms."