Google Cloud: AI, a Security Challenge for Councils

Google Cloud's Chief Operating Officer, Francis de Souza, recently emphasized the critical importance of integrating security into companies' artificial intelligence (AI) strategies from the outset. At an event in Los Angeles, he stated that an AI strategy cannot exist without a corresponding data and security strategy. According to TechCrunch, de Souza also warned about the risks posed by "shadow" AI, where employees use AI tools without adequate oversight from the company. He also mentioned the danger of AI agents accessing forgotten data sources, such as old SharePoint servers.

The security of AI systems now extends beyond traditional network boundaries. "In addition to your usual infrastructure, you now have models, data pipelines for training models, agents, and prompts. All of this requires adequate protection," he said. De Souza highlighted that the time between an initial breach and the next stage of an attack has drastically decreased, from eight hours to just 22 seconds. To address these threats, he advocates for an agent-based defense, where humans oversee a fully automated defense managed by agents.

De Souza stressed that security cannot be an afterthought, nor can it be left to the discretion of employees. Companies must adopt a unified security strategy that covers all clouds and models, even if they believe they are using only one provider. Indeed, SaaS applications and business partners often introduce multiple clouds into the equation, making a comprehensive approach essential. According to him, AI security must be discussed at the boardroom level, and not just viewed as an IT department issue.