Dutch Startup Launches Open-Source Code to Make AI Reliable
Amsterdam, Thursday 13 August 2026
Dutch startup LemmaBase has launched Lemma, an open-source programming language designed to bring absolute reliability and legal certainty to AI systems by translating complex regulations into verifiable code.
The Growing Trust Deficit in Modern Enterprise AI
As organisations globally rush to integrate artificial intelligence into their core operations, a widening gap between enthusiasm and execution has emerged. According to a study by software firm Workiva, approximately 25 per cent of Dutch organisations have discovered AI-generated errors during internal audits that had already reached external stakeholders or the boardroom [5]. This rate aligns closely with the global average of 26 per cent, representing a minimal difference of just 1 percentage point [5]. Despite these slip-ups, 81 per cent of Dutch professionals maintain some level of trust in AI’s accuracy without requiring constant human oversight [5]. However, this optimism clashes with a stark reality: only 14 per cent of those surveyed believe their internal data is sufficiently reliable for large-scale AI deployment [5].
Bridging the Gap with ‘Rules as Code’
To address the inherent unpredictability of probabilistic generative AI models, Amsterdam-based startup LemmaBase officially launched its open-source programming language, Lemma, on 10 August 2026 [1][4]. Designed as a direct counterweight to the closed-loop, black-box systems favoured by global tech giants, Lemma introduces the concept of ‘rules as code’ [1][2][3]. Unlike traditional generative models that calculate outputs based on probabilities, Lemma translates complex legislation, business processes, and product terms into deterministic code [1][3]. This approach ensures that computer systems execute guidelines with absolute certainty, making the underlying logic fully transparent, auditable, and verifiable by both humans and machines [1][2][3].
Deterministic Utility Across Regulated Sectors
The practical applications of Lemma are tailored for highly regulated environments where operational errors carry severe legal or financial consequences [1][2]. In the financial sector, for instance, a conversational generative AI agent can manage customer interactions for a mortgage application, while Lemma operates in the background to calculate borrowing capacity strictly according to current statutory guidelines [1][3]. Similarly, public sector institutions can encode complex tax regulations and welfare eligibility criteria into Lemma, eliminating administrative arbitrariness and ensuring decisions are fully explainable [1][3]. Municipalities can also deploy the language to automate compliance checks for building permit applications, while insurers and logistics companies can use it to guarantee flawless execution of policy payouts and international customs rules [1][3].
A Sovereign Move Towards Transparent AI Governance
The release of Lemma’s source code and specifications on GitHub and LemmaBase.com marks a pivotal shift from raw AI capability to structured AI governance [1][2][3]. By decoupling decision-making logic from underlying operational systems, organisations can leverage AI’s speed without relinquishing regulatory control [1]. This open-source framework allows developers to audit the software’s inner workings, offering a transparent alternative to the proprietary models of dominant American technology firms [2][3]. As European regulatory bodies increasingly demand explainable and trustworthy AI systems, Lemma provides a practical mechanism for companies to achieve compliance and maintain sovereign control over their digital infrastructure [1][2].