Thales and Google Cloud expand partnership to secure AI agents

By Splitfeed EditorialLast updated Sep 29, 2026, 11:30 UTC

On September 28, 2026, Thales and Google Cloud announced an expanded partnership aimed at securing agentic AI workflows. The collaboration centers on governance and security controls for AI agents, including those operating on Google's Gemini platform. The announcement was distributed through Business Wire and picked up the same day across financial and technology publications including Yahoo Finance, GuruFocus, and TradingView.

The partnership matters because agentic AI, systems that act autonomously on behalf of users or organizations, raises a distinct set of security questions beyond those posed by conventional AI tools. Unlike a model that simply generates text, an AI agent can query databases, trigger actions, and communicate with other systems. Who or what it can reach, what data it can read or share, and whether its actions can be audited are live governance problems that enterprises are only beginning to tackle. The Thales involvement is notable because the company is a major player in data security and encryption for large organizations, suggesting the deal is oriented toward enterprise customers with existing compliance obligations rather than consumer-facing products.

The pro-AI camp has not commented on this story yet, though stories of this type typically draw the argument that building security infrastructure around AI agents is exactly the responsible path forward, and that partnerships between specialized security firms and cloud providers demonstrate the industry taking risk management seriously. The middle ground camp has framed the deal in broadly positive terms, with Stateside Wire noting that the integration introduces protections governing what agents can access, share, and do, and ttakanas.bsky.social describing security and governance as foundational layers rather than afterthoughts. The anti-AI camp is more skeptical, with Blah blah blah arguing that tech companies have not managed to contain existing threats like spam, bots, and fake accounts, making further AI deployment hard to justify, and Flingjore pointing to the asymmetry that attackers can now conduct large-scale breaches cheaply, which any new security layer will have to contend with.

The practical test for this partnership will come as enterprises begin deploying Gemini-based agents at scale. Specific disclosures about what governance controls are implemented, whether they are auditable by third parties, and how they perform against adversarial use cases would give observers something concrete to evaluate beyond the announcement itself.