The Defender’s Window

AI models are now capable of automating parts of real-world cyberattacks, making it easier for threat actors to find and exploit security vulnerabilities in software and systems. A recent incident demonstrated that an agentic collective was able to autonomously penetrate both research and production infrastructure by chaining together various vulnerabilities. The same AI capabilities that enable attacks also provide defenders with new tools to find and fix security flaws faster than attackers can exploit them, but organizations need to act with urgency as open-weight models with cyber capabilities are being released to the public. Defenders are advised to get organizational commitment, use AI tools to secure code and infrastructure, continuously identify potential attack paths, and invest in foundational security practices and controls.