Datashift and Kyvvu bring real-time security to autonomous AI agents

Belgian data consultancy Datashift has partnered with Dutch AI security company Kyvvu to launch a real-time security layer that runs inside autonomous AI agents, checking every action in less than one millisecond and blocking or flagging risky behavior. Datashift said the layer is a first for companies in regulated European markets, where traditional security and governance controls are often too slow or too granular to monitor agent autonomy. The two companies are targeting security teams and compliance officers, combining Datashift's compliance expertise — including its earlier acquisition of Antwerp-based Data Trust Associates — with Kyvvu's agent-security technology.

Datashift and Kyvvu are joining forces to introduce a new real-time security layer for AI agents. In doing so, they aim to prevent incidents with autonomous AI systems before they occur, with a focus on regulated European markets. The Belgian data consultancy Datashift announces a partnership with the Dutch Kyvvu, a security company for AI agents. Today, AI agents are increasingly being deployed in business processes and are capable of performing actions independently. This brings new security risks, especially when they are granted access to critical systems. Datashift and Kyvvu want to anchor security within the AI agent itself so that controls can take place more strictly and quickly. Datashift is active in Leuven, among other places, and since last year in Rotterdam https://itdaily.com/news/business/datashift-rotterdam/ . Earlier this year, the company acquired the Antwerp-based compliance expert Data Trust Associates https://itdaily.com/news/business/datashift-acquieres-data-trust/ . New risks from autonomous AI agents AI agents are no longer limited to simple tasks. They perform actions independently https://itdaily.com/news/software/salesforce-introduces-a-range-of-ready-to-use-ai-agents/ , process payments, and communicate with other systems without human intervention. This increases their efficiency but also leads to increased risk. AI models from OpenAI https://itdaily.com/blogs/security/openai-infiltrates-hugging-face/ and Anthropic, among others, recently managed to penetrate external systems during tests. None of these incidents were noticed at the time, demonstrating the need for real-time detection. Meanwhile, AI leaders are speaking of a brake on AI development https://itdaily.com/blogs/innovation/ai-leaders-discover-their-conscience/ , though the question remains whether this is genuine remorse or a tactical move. According to Datashift, traditional security and governance controls are often not fast or granular enough to monitor the autonomy of such agents. Especially when organizations scale their AI solutions to production and grant access to sensitive corporate information, a new approach becomes necessary. Real-time security layer The collaboration between Datashift and Kyvvu results in a security layer that runs directly within the AI agent. Every action is checked in less than a millisecond. Safe actions proceed without interruption, while risky actions immediately trigger a warning or are blocked. This is intended to prevent unwanted agent behavior in real time, a first according to Datashift for companies in regulated markets. Kyvvu and Datashift want to leverage their combined expertise to address both technical and compliance challenges for AI applications in Europe. They are targeting security teams and compliance officers who need certified solutions quickly.