M E N U C L O S E

AI Charter

At Weichie, we believe in the power of innovation, but always under human supervision. This charter describes how we deploy Artificial Intelligence (AI) to optimize our development processes without compromising on quality, security, or transparency.

In a digital agency, transparency, human control, and accountability are essential when using artificial intelligence. With this charter, Weichie.com commits to guaranteeing these principles at all times. Driven by an interest in innovation, we look at the opportunities AI offers with great enthusiasm. We are also fully aware of the potential pitfalls and risks that AI entails. Therefore, this charter outlines the framework within which we use AI at Weichie.

1. Human Oversight & Responsibility

For us, AI is a co-pilot, not the captain.

  • No Live Deployments: AI does not have direct write access to our production environments. All code always passes through our standard CI/CD pipeline (e.g., via Bitbucket), including manual code reviews.
  • Ethics: Our developers remain personally responsible for the delivered quality, in accordance with our internal coding standards and ethical principles.
  • Final Review: Every AI-suggested change is reviewed and validated by a senior developer for logic, security, and maintainability. We never deploy code that we do not 100% fully understand ourselves.
  • OBO authentication: Our AI agents act On-Behalf-Of (OBO) our real developers. It creates a secure way for agents to operate with scoped, revocable permissions, while maintaining a clear audit trail that says exactly who acted, on whose behalf, and within which scopes. This ensures accountability, traceability, and rapid revocation if misuse is detected.

2. Privacy, Security & IP

The protection of our clients’ data and intellectual property is our highest priority.

  • No Sensitive Data: Personal data (GDPR) or secret API keys are never shared with AI models.
  • Security Scans: AI-generated code undergoes extra checks for security vulnerabilities to ensure the integrity of the application.
  • Data Isolation: We exclusively use AI tools (such as Claude for Business) where the entered data is not used to train public models.
  • IP Ownership: All AI-assisted work produced for clients is contractually owned by Weichie or the client, exactly as if it were written entirely by a human developer.

3. Transparency towards the Client

We are honest about the role of AI in our creative process.

  • Content & Media: For AI-generated text, images, or videos visible to the end user, we maintain a duty of transparency. If content is *substantially generated by AI, we will state this clearly.
  • Source Material: We prefer using the client’s own material or licensed content (CC0).
  • Development: Because AI serves purely as an efficiency tool in development and every line is validated by a human, we do not disclose this per individual line of code.
  • Client Preference: Clients may request AI-assisted or AI-limited workflows for their projects. These preferences are respected and documented.

* Substantially generated refers to content where AI produces the majority of the final output without significant human rewriting.

4. Quality Assurance (QA)

We are aware of potential “AI slop” and low-quality generated output. Therefore, all our code is stored in a version-control system and reviewed before final launch.

  • Final Accountability: Weichie takes full responsibility for any code created with or without AI assistance.
  • Uptime Guarantee: Reliability is non-negotiable. The use of AI must never lead to instability. Weichie remains fully liable for the uptime and performance of the projects we manage.
  • Coding Standards: Our AI tools are guided by our own internal codebase and coding standards. This ensures the output aligns seamlessly with our architecture.
  • Testing: AI-generated code is subject to the same automated testing, staging validation, and rollback procedures as human-written code.

The ultimate responsibility for the quality and uptime of our website projects lies entirely with Weichie.

5. Bias, Hallucinations & Model Limitations

We are aware that AI is not a one-shot guarantee.

  • AI systems can produce incorrect, biased, or outdated information.
  • We never rely on AI for factual, legal, financial, or security-critical decisions without independent human verification.
  • Outputs are treated as suggestions, not truths.

6. AI Vendor & Model Governance

We only use AI vendors that meet our security, privacy, and compliance standards. This can be a combination of multiple models and AI agents based on the scope of work.

  • Experimental models are never used on client production projects without validation.
  • We align our AI usage with applicable laws and regulations, including GDPR and the EU AI Act.
  • High-risk use cases are explicitly excluded from our workflows unless legally and ethically justified.