technology
Kyndryl unveils policy as code to constrain agentic AI
Kyndryl has launched a policy as code process that converts organisational rules into machine-readable guardrails to prevent agentic drift and help enterprises meet rising compliance demands ahead of new EU rules.
Apr 3rd 2026 · United States
Insights
- Kyndryl announced a policy as code capability last month to encode organisational rules into machine-readable guardrails.
- Kyndryl’s Readiness Report found 31% of enterprise customers cite regulatory or compliance concerns as a main barrier to scaling new technology investments.
- Policy as code translates policies into executable instructions so AI agents operate only within pre-defined limits while human experts retain oversight.
- Kyndryl says the approach reduces the risk of agentic drift by constraining how agents reason, adapt and act at runtime.
- The company argues policy as code improves transparency, explainability, lowers operational risk and speeds AI-native workflows in regulated sectors such as finance, public services and supply chains.
- EU AI Act transparency rules take effect in August 2026, increasing compliance pressure on enterprises to adopt governance measures.
Sources
- Group Pushing Age Verification Requirements For AI Sneakily Backed By OpenAI news.slashdot.org
- OpenAI shutters controversial AI video generator Sora www.siliconrepublic.com
- Policy as code: Embedding compliance in AI adoption www.siliconrepublic.com
- Dubious AI detectors drive ‘pay-to-humanise’ scam www.dawn.com
- Beijing mandates internal AI ethics reviews to ensure ‘controllable’ tech www.scmp.com