Everyone is navigating AI security in real time — even Google

Google Cloud’s Chief Operating Officer, Francis de Souza, highlighted the urgent need for companies to integrate security into their AI strategies from the outset. Speaking backstage at an event in Los Angeles, de Souza emphasized that security cannot be an afterthought or left solely to employees, especially as organizations adopt AI technologies. He warned against “shadow AI,” where employees use consumer AI tools without oversight, urging businesses to demand governance, auditability, and robust security measures as part of any AI platform. De Souza stressed that AI strategies must be closely aligned with data and security strategies, forming a cohesive approach rather than isolated efforts. He also advocated for a multicloud security posture, noting that most companies operate across multiple cloud environments, even if they believe they rely on a single provider. This complexity requires consistent security policies that span cloud platforms and service models to effectively manage risks. The evolving threat landscape, according to de Souza, demands new defensive models that operate at machine speed. He pointed out that the average time between an initial breach and subsequent attack stages has shrunk dramatically—from eight hours to just 22 seconds—making traditional security approaches inadequate. The attack surface now includes not only networks but also AI models, data pipelines, and prompts, all of which must be secured. A particularly overlooked risk involves AI agents navigating internal systems, potentially uncovering forgotten or poorly secured data repositories such as outdated SharePoint servers. These hidden assets can expose sensitive information if not properly managed. De Souza advocates for AI-native, agent-driven defenses that can respond as quickly as threats emerge, signaling a shift toward automated, adaptive security frameworks designed for the complexities of AI-enabled environments.