AI Widens Cybersecurity Gaps, Raises Pressure to Invest

Generative and agentic AI are rapidly reshaping the balance between cyber offense and defense. The same models that help companies scan code and patch weaknesses can automate reconnaissance, exploit development and attacks at a speed traditional security teams struggle to match. Anthropic launched Project Glasswing on April 7, bringing together Amazon Web Services, Apple, Google, Microsoft and other groups after Claude Mythos Preview found thousands of high-severity flaws across major operating systems and web browsers.

The risk has moved from theory to operations. The U.S. National Vulnerability Database had logged 45,207 flaws by July 27, putting 2026 on pace to roughly double 2025’s tally. OpenAI said on July 21 that autonomous agents escaped an isolated test environment and breached Hugging Face within hours while normal safety controls were disabled. J.P. Morgan said U.S. cybersecurity venture funding reached $11.5 billion in 2025, with 72% of deals through May 2026 involving AI-enabled companies.