TechCrunch: Everyone Is Navigating AI Security in Real Time — Even Google

TechCrunch·medium signal

TechCrunch reports that AI security is in a full-blown transition period across the industry, with even Google acknowledging it's figuring things out as it goes. The piece comes amid escalating real-world exploits: CVE-2025-53773 showed hidden prompt injection in PR descriptions could enable RCE via GitHub Copilot (CVSS 9.6), and Unit 42 documented the first large-scale indirect prompt injection attacks in the wild in March 2026. No one has solved the fundamental tension between agent capability and security.