← Back to Tsalach Ventures

AI UNDER HUMAN RESPONSIBILITY

We build with AI. We never delegate our responsibility to it.

AI accelerates our research, design, development and operations. Its use remains governed by human rules proportionate to the level of risk in each product.

  1. Human accountability

    An identifiable person remains accountable for every material decision, production release and business consequence.

  2. Risk-proportionate verification

    The greater a feature’s potential impact on a user’s finances, safety, rights or reputation, the stronger its human and technical controls must be.

  3. Meaningful transparency

    We clearly disclose when a user is interacting with AI or when a material output has been generated or recommended by AI.

  4. Data protection and minimization

    We collect only what is necessary, restrict access and do not send sensitive data to a model without a legitimate purpose and appropriate safeguards.

  5. Security by design

    Access, secrets, logs, data and automated actions are protected according to system risk. An AI system never receives more permissions than it needs.

  6. Context and fairness

    We test products across the languages, cultures and real operating conditions of the markets they serve to reduce imported errors, exclusions and bias.

  7. Resilience and independence

    We avoid leaving critical functions dependent on a single model, provider or automated agent without a practical fallback.

  8. Continuous monitoring

    Systems are monitored after launch. Incidents, errors and user feedback are used to improve the product and strengthen its safeguards.

Our goal is not maximum automation. It is reliable, understandable and durable value.