AI Transparency

Last updated: 23 February 2026

In accordance with Article 50 of the EU AI Act, this page provides transparent information about how Pitchwits uses artificial intelligence in its grant-writing platform.

1. How We Use AI

Pitchwits uses AI across three core functions of the platform:

1.1 Content generation

AI generates draft sections of grant applications based on your company information, questionnaire responses, and uploaded documents. Each section is generated individually, allowing you to review and edit content incrementally. You can regenerate sections or specific paragraphs at any time.

1.2 Quality assessment

After generating a draft, AI assesses the quality of the output by comparing it against source documents and evaluating completeness, specificity, and alignment with grant programme requirements. This produces a feedback report highlighting strengths and areas for improvement, along with a completeness score.

1.3 Document extraction

When you upload documents (pitch decks, business plans, technical papers), AI extracts structured information such as company details, technology descriptions, market data, and financial figures. This extracted data is used to inform content generation.

2. AI Models

We use the following AI models:

All models are accessed via paid API endpoints. Neither Anthropic nor Google use data submitted via their paid APIs to train their foundation models.

3. Human Oversight

AI outputs go through a multi-stage human oversight process before any use in a grant submission:

1. Draft generation — AI produces a draft section based on your inputs and uploaded documents.
2. Automated quality assessment — A second AI model evaluates the draft for completeness, accuracy, and alignment, producing a feedback report with specific improvement suggestions.
3. Human review — You review the draft and assessment feedback. You can accept, edit, or fully regenerate the content.
4. Iterative refinement — You can regenerate specific paragraphs, request reassessment, or make manual edits until satisfied.
5. Final approval — No content is exported or submitted without your explicit action. The final document reflects your editorial decisions.

At every stage, you retain full control. AI is an assistive tool; it does not make decisions or take actions autonomously.

4. Limitations and Accuracy

AI-generated content has inherent limitations that you should be aware of:

• Accuracy — AI may produce content that is factually incorrect, internally inconsistent, or misaligned with your specific situation. Always verify factual claims.
• Hallucination — AI models can generate plausible-sounding but fabricated details, particularly for specific figures, dates, or technical specifications not present in the source material.
• Placeholder content — When the AI lacks sufficient information, it may generate generic content that needs to be replaced with your specific details.
• Bias — AI models may reflect biases present in their training data. Review generated content for unintended bias or inappropriate assumptions.
• Not a guarantee — AI-generated content does not guarantee funding success. Grant outcomes depend on many factors beyond the quality of the written application.

5. EU AI Act Classification

Under the EU AI Act, our AI system is classified as limited-risk (Article 50 transparency obligations apply). It is a content-generation tool that:

• Assists humans in drafting text, with full human editorial control
• Does not make autonomous decisions affecting individuals' rights
• Does not assess creditworthiness or eligibility
• Does not perform biometric identification or categorisation
• Does not fall under any high-risk use case defined in Annex III

Our system is comparable to general-purpose AI assistants used for content creation, and we comply with the transparency requirements of Article 50 by clearly disclosing AI involvement.

6. Data Protection

Your data is processed in accordance with GDPR. Key safeguards include:

• Data sent to AI providers is processed under their data processing agreements with Standard Contractual Clauses (SCCs)
• Training data consent is strictly opt-in and can be withdrawn at any time
• Any data used for model improvement is anonymised before processing
• You can download all your data or request account deletion from your Privacy Settings

For full details, see our Privacy Policy.

7. Contact for AI Concerns

If you have questions, concerns, or complaints about our use of AI, contact us at:

See also our Terms of Service and Privacy Policy.