Datashift and Kyvvu bring real-time security to autonomous AI agents
IT Daily
4 min read
16 Sep 26
Datashift is partnering with Kyvvu, a Dutch company specializing in real-time control and governance for AI agents.
Together, we're adding a control layer to agentic solutions that evaluates actions in real time and enforces centrally defined policies before those actions are executed. This adds an additional level of protection on top of the permissions, access controls and guardrails already built into the agents themselves.
The combination brings together Datashift's experience in designing, building and deploying AI agents in real business environments with Kyvvu's technology for consistently enforcing controls across the actions those agents take.
Why does this matter?
A well-designed enterprise agent does not simply get unrestricted access to systems and start making decisions on its own. Its tools, permissions, data access and actions are deliberately configured and constrained.
The challenge starts when you scale.
Agents evolve. New tools and integrations are added. Permissions change. Different teams modify workflows. Agents start calling other agents. And despite good engineering practices, not every possible combination and edge case will always be tested end-to-end after every change.
At that point, the main risk is often much more mundane than an AI model “going rogue”: a configuration mistake, an overly broad permission, a forgotten edge case or simply human error.
That is why a central control layer becomes necessary.
Instead of relying on every individual agent and integration remaining perfectly configured, organisations can define critical security and governance policies once and enforce them consistently across their agent ecosystem, in real time.
Think of it as the seatbelt for agentic AI. Not because you expect every journey to end in an accident, but because once you operate at scale, you need a safety mechanism that does not depend on every person, configuration and implementation being perfect.
That is the type of infrastructure we believe organisations will need as agents move from isolated experiments to production at scale.
Full story by ITdaily: https://lnkd.in/eAaA6wNx
#AIAgents #AISecurity #AIGovernance
Keep reading
Eager to learn more? No worries: we’ve got you covered.
Fair Beds, Not a Race: How Datashift Built a Smarter Way to Allocate Night Shelter in Rivierenland
Datashift developed a digital application for CAW Rivierenland that allocates the available beds automatically, transparently and fairly across twelve local social welfare services (OCMW's). The project was covered by HLN.
Kris D'hulst in Trends | What's left of us after AI?
Our Operations Director, Kris D'Hulst, took to Trends DataNews with a question the AI debate tends to skip right past. Not what AI means for the economy or the job market but what makes each of us irreplaceable as individuals, once machines take over more of our thinking and knowledge work. It's a question close to how we think about adopting AI at Datashift, so we're proud to share his take.
Datashift welcomes Ilias (18) as "Boss of Tomorrow"
As part of JINC's "Boss of Tomorrow" initiative, 18-year-old Mechelen student Ilias spent a day shadowing Datashift CEO Nico Huybrechts. The programme, run by equal-opportunity organisation JINC, pairs pupils and teenagers with company leaders to help them explore career paths and give business leaders insight into how the next generation views their roles.
