By IddiLabs
A Risk Manager's journey into AI. Articles, real experiments, practical tools and honest lessons from the compliance & operations frontline.
Small PSFs can deploy AI in 30 days while big banks are still in procurement. Limited budgets force focus. Small teams move fast. Open-source models run locally. Constraints become advantages.
AI models drift over time and change via vendor updates. Risk managers need version-controlled, static models to ensure reproducible compliance reports—only possible with self-hosted open-source AI.
AI SaaS lock-in is a compliance risk. When vendors change pricing or deprecate models, your workflows break. Open-source models eliminate dependency—you own the asset forever.
AI vendors are now critical ICT dependencies under DORA. Closed-model APIs create concentration, outage, and supply-chain risks. If you haven’t assessed your AI provider, you’re already exposed.
Closed AI models can’t meet the EU AI Act’s transparency and documentation rules. If you can’t explain the model, you can’t deploy it—open-weights models are the only viable path for compliance.