Introduction

FISCAL Technologies is committed to responsible AI development and use in our business solutions for finance teams. This statement outlines our position regarding the European Union’s Artificial Intelligence Act (EU AI Act) following a comprehensive evaluation of our solutions against the requirements established by this legislation.

Our Assessment Results

FISCAL Technologies has determined that while we qualify as an AI provider under the Act, our solutions do not carry mandatory obligations under the EU AI Act’s regulatory framework. This determination is based on the following specific factors:

1. Purpose and Function: Our AI systems analyse invoice data to identify potential risks in payments that have already been initiated by customers. Our technology serves purely in an advisory capacity.
2. No Personal Data Processing: Our AI systems do not process personal identifiable information (PII).
3. No Automated Decision-Making: Our solutions do not make autonomous decisions affecting individuals – they only highlight potential risks in invoices where a payment decision has already been made by the customer.
4. Risk Classification: Following thorough assessment against the EU AI Act’s risk-based approach, our solutions do not fall under prohibited or high-risk categories.

Risk Classification Analysis

Prohibited AI Applications

We confirm that FISCAL Technologies does not develop or deploy any AI systems that fall into the prohibited categories as defined by the EU AI Act, including systems that:

• Deploy subliminal or manipulative techniques
• Exploit vulnerabilities of specific groups
• Implement social scoring systems
• Conduct real-time remote biometric identification in publicly accessible spaces for law enforcement purposes

High-Risk AI Systems

Our solutions do not qualify as high-risk AI systems as they:

• Do not serve as safety components of products
• Are not used in critical infrastructure management
• Do not determine access to educational or vocational training
• Are not used for employment, worker management, or access to self-employment
• Do not evaluate creditworthiness or establish credit scores of individuals
• Are not used in the administration of justice or democratic processes

General-Purpose and Non-High-Risk AI Systems

Our AI-driven solutions fall into the category of non-high-risk AI systems, and as such, do not carry mandatory obligations under the EU AI Act.

Our Voluntary Commitments

While exempt from mandatory obligations, FISCAL Technologies embraces the spirit of the EU AI Act and voluntarily commits to:

• Following the voluntary code of conduct outlined in Article 95 of the EU AI Act
• Maintaining appropriate levels of AI literacy among our employees in alignment with Article 4
• Ensuring transparency in how our AI systems operate
• Maintaining robust data governance and quality standards
• Regularly testing and validating our AI models for accuracy and reliability
• Making clear information available to our customers about how our AI systems function
• Implementing appropriate security measures to protect the integrity of our systems

Ongoing Governance

FISCAL Technologies has established internal governance processes to:

• Monitor changes in AI regulatory requirements
• Conduct regular assessments of our AI systems
• Update our AI governance framework as needed
• Train staff on responsible AI practices

3^rd Party AI

Below is a list of third-party AI utilised within FISCAL Technologies products.

Azure OpenAI Service

Our integration with the Azure OpenAI Service is subject to the following disclosures in compliance with the EU AI Act:

1. AI System Provider
   This feature uses generative AI technology developed by Microsoft.
2. Purpose and Functionality
   The Azure OpenAI integration is used to process user requests for the creation of graphical reports and convert them into our query structure so they can be executed as normal database retrieval operations. The user reviews the resulting design before manually accepting or rejecting it.
3. Data Usage
   Customer inputs processed through this AI system are handled in accordance with our Privacy Policy and Microsoft’s data processing terms. No customer data is used to train the underlying models without explicit consent.
4. Limitations
   While Azure OpenAI provides valuable functionality, it has inherent limitations, including potential for inaccuracies in responses. All AI-generated outputs should be reviewed by appropriate personnel before being used for decision-making purposes.
5. Risk Mitigation
   FISCAL Technologies implements appropriate safeguards including never giving the AI direct access to customer data and only allowing it to make edits to a limited query structure that the user reviews before accepting, to ensure responsible use of this technology.

Azure Computer Vision Service

Our integration with the Azure Computer Vision Service is subject to the following disclosures in compliance with the EU AI Act:

1. AI System Provider
   This feature uses deep learning models developed by Microsoft, specifically trained for text detection and recognition. These OCR (Optical Character Recognition) models are discriminative models trained to recognize patterns in images that correspond to text, rather than generative AI models designed to create new content.
2. Purpose and Functionality
   The Azure Computer Vision integration is used to extract positioned text from images and PDFs to enhance document processing capabilities within our software.
3. Data Usage
   Customer inputs processed through this AI system are handled in accordance with our Privacy Policy and Microsoft’s data processing terms. No customer data is used to train the underlying models without explicit consent.
4. Limitations
   While Azure Computer Vision provides valuable functionality, it has inherent limitations, including potential for inaccuracies in text recognition, particularly with:
   1. Low-quality images or scans
   2. Unusual fonts or handwriting
   3. Documents with complex layouts

All AI-extracted text should be reviewed by appropriate personnel before being used for decision-making purposes.

5. Risk Mitigation
   FISCAL Technologies implements appropriate safeguards including:
   1. Requiring user review and approval before accepting any AI-processed results
   2. Implementing quality control measures to identify potential recognition errors

Contact Information

For questions regarding our EU AI Act compliance or to report concerns, please contact: [email protected]

This statement confirms FISCAL Technologies’ commitment to responsible AI development and use while clarifying our specific position in relation to the EU AI Act.