The global AI market is on track to reach $1.3 trillion by 2030, per Statista’s 2024 data—but rapid adoption is bringing growing security risks. A 2024 McKinsey report on enterprise AI found 60% of employees use unapproved AI tools (known as shadow AI), leaving companies open to data leaks and compliance issues. It was in this context that Google Cloud COO Francis de Souza shared his take on AI security during a backstage interview at a Los Angeles event.
In a calm, thoughtful tone, de Souza emphasized security can’t be an afterthought. “As companies set out on their AI journey, they need a platform approach,” he explained, noting security should be built into systems from the start rather than tacked on later. He warned shadow AI—when employees use consumer tools like ChatGPT without oversight—increases breach risks threefold, citing Gartner’s 2023 findings. To fix this, enterprises must demand AI platforms with security, governance, and auditability features, ensuring every interaction is traceable and GDPR-compliant.
Even Google, an AI leader, is working through these challenges in real time. De Souza pointed to Google Cloud’s Vertex AI platform, which includes end-to-end encryption for data at rest and in transit, role-based access control (RBAC), and audit logs tracking model usage. These features cut shadow AI risks by offering a secure alternative to consumer tools: for instance, Vertex AI’s Model Garden lets organizations deploy pre-trained models with built-in governance, so sensitive data doesn’t end up in unvetted tools.
The risks are significant: GDPR fines for breaches can hit 4% of annual global revenue, and IBM’s 2024 report found AI-related breaches cost an average $4.5 million—15% more than non-AI breaches. De Souza’s guidance drives home that security has to be a core part of any AI strategy, not an optional extra.
Google Cloud isn’t alone in focusing on AI security. Microsoft Azure’s AI Studio has ISO 27001 and GDPR-aligned compliance tools, while AWS Bedrock offers model governance features like versioning and access controls. As companies prioritize security when picking vendors, these platforms are competing to deliver the most secure solutions.
Recent industry moves show the White House’s 2023 AI Security Executive Order has ramped up attention on AI safety, with 2024 guidelines requiring vendors to share security test results with federal agencies. This regulatory pressure is pushing tech giants and enterprises to tighten their AI security practices, ensuring systems are both safe and reliable.
快报