Use of Artificial Intelligence and Transparency

1. How ScoringMy uses AI

ScoringMy's artificial intelligence features are used to support professional communication and provide analysis tools. The providers used have distinct roles.

Anthropic Claude · text generation and adaptation

ScoringMy uses Anthropic Claude to generate and adapt professional text content.

When users work with these features, ScoringMy may provide the model with the professional and communication context needed to generate relevant content. Where configured by the user, this context may include: professional positioning, communication topics and approaches, areas in which they want to build visibility, subject areas, keywords, communication style, language and value proposition.

Where a client organisation has defined specific guidelines, communication red lines may also be used to guide content generation.

ScoringMy applies data minimisation and purpose limitation, using the context required for the requested feature.

Google Gemini · image generation

ScoringMy uses Google Gemini to generate images intended to accompany professional content.

Images are generated using the textual content of the post as context, so that the resulting image relates to the content the user is preparing. Generated images must be reviewed by the user before use.

OpenAI · semantic analysis of posts

ScoringMy uses OpenAI to perform semantic analysis of the textual content of certain professional posts.

For this analysis, ScoringMy sends OpenAI only the text of the post being analysed. The request does not include, as associated metadata, the employee's name, email address, professional profile URL, author identity or other direct identifiers associated with the user.

Semantic analysis may be used to:

• Categorise content, for example as corporate, personal, educational or informational.
• Detect references or mentions of the company.
• Assess content against red lines or communication restrictions previously defined by the client company.
• Classify posts within communication territories defined by the organisation.

2. AI providers and data used

Provider	Purpose	Context / information used
Anthropic Claude	Text generation and adaptation	Instructions and professional/communication context configured where required for the feature; client-defined red lines may be included.
Google Gemini	Image generation	Textual content of the post used as context to generate a related image.
OpenAI	Semantic analysis	Post text. No name, email, profile URL, author or other identifying metadata is attached to the request.

AI integrations are performed from the ScoringMy backend through APIs. API credentials are not exposed to the user's browser.

PeopleXBrand does not use prompts, user-provided content or outputs generated through ScoringMy to train or retrain models of its own. Any processing carried out by AI providers is governed by the applicable contractual, privacy and API service terms.

3. Professional news and topics

ScoringMy may suggest publicly available professional news and industry-related topics based on the user's professional context, to help them identify relevant subjects and prepare draft professional posts.

Suggestions and the resulting content are editable and require human review before use or publication. Users must verify the source, date, context and accuracy of the information when content is based on a news item.

4. Human oversight

AI-generated text and images can be reviewed, edited and adapted by the user before use.

ScoringMy does not automatically publish AI-generated content.

The final decision on whether content should be used or published rests with the user.

5. Employee communication analytics

ScoringMy may analyse professional content and generate indicators relating to communication, activity and positioning. Semantic analysis outputs — including categorisations, company references, red lines and communication territories — should be understood as indicators supporting communication analysis.

AI systems may misinterpret context, produce incorrect classifications or generate false positives or false negatives. Their outputs therefore require contextual interpretation and human oversight.

6. AI limitations

Artificial intelligence systems are not infallible. AI-assisted content and analysis may contain errors, incomplete interpretations, outdated information, stylistic variations, bias or different results depending on context.

In particular, errors may occur in content classification, identification of indirect references to a company, interpretation of red lines, assignment of communication territories or the relationship between a generated image and the meaning of the text.

ScoringMy therefore combines AI features with human oversight, application controls, structured instruction design and user control over publication.

7. Responsible use

Users of AI features must use ScoringMy responsibly and in accordance with the law and their organisation's internal policies.

Before publishing AI-assisted content, the user should:

• review the factual accuracy of the content;
• verify sources, dates and statistics where applicable;
• review the appropriateness of tone and professional context;
• check that generated images are appropriate and not misleading;
• avoid confidential information, special categories of data or unnecessary personal data;
• confirm compliance with their organisation's communication policies, red lines and guidelines.

AI-generated content is a support tool and does not replace the user's professional responsibility.

8. Transparency and regulatory framework

ScoringMy applies principles of transparency, human oversight, data minimisation and responsible use to its AI-assisted features.

ScoringMy integrates third-party models through APIs and does not develop or distribute foundation models of its own. The legal classification and obligations applicable to each feature depend on its purpose and the specific deployment and use context.

9. AI documentation

For further information on the architecture, risks, limitations and responsible use of artificial intelligence in ScoringMy:

9.1. AI System Description

AI architecture, providers, features, data flows and human oversight.

1. Overview of the System

ScoringMy is a SaaS platform designed to help professionals and organizations improve digital professional presence through metrics, insights, communication analytics and AI-assisted content creation. ScoringMy does not develop or train foundation models; it integrates third-party AI services through secure APIs.

2. AI System Architecture

AI capabilities are orchestrated by the ScoringMy backend. The web application communicates with backend business logic, which prepares controlled requests and invokes third-party AI services server-side. AI API credentials are not exposed to end users.

3. AI Model Providers and Purpose

• Anthropic Claude: generation and adaptation of professional text content.
• Google Gemini: generation of images intended to accompany professional posts.
• OpenAI: semantic analysis of post text for content classification and communication analytics.

4. Data Flow by Functionality

4.1 Text generation with Claude. The user requests or configures a professional content draft. The backend may construct the request using professional and communication context optionally supplied by the user, including positioning goals, areas in which the user wants to build visibility, preferred topics and approaches, keywords, communication style, language and value proposition. Where configured by the client, communication red lines may also be included to constrain the output.

4.2 Image generation with Gemini. After text content for a post has been generated or prepared, the platform may use that textual content as context to request an image that is relevant to and suitable for accompanying the post.

4.3 Semantic analysis with OpenAI. For semantic analysis, the backend sends only the text of the post being analyzed. ScoringMy does not send the associated employee name, email address, profile URL, author identity or other identifying metadata as part of this analysis request.

The returned analysis may be used to:

• classify posts into content categories, including corporate, personal, educational or informational
• identify whether the post references or mentions the company
• assess alignment with client-defined communication red lines
• classify the post within communication territories defined by the client.

5. Data Minimization

The data sent to each AI provider is limited to the information required for the corresponding functionality. ScoringMy avoids sending direct identifiers when they are not necessary for the requested operation and instructs users not to introduce sensitive personal data or unnecessary confidential information.

6. AI Functionalities

6.1 Content Generation. Claude may generate suggested professional posts, drafts and adaptations. Outputs are editable and intended to assist users in creating professional content.

6.2 Image Generation. Gemini may generate images based on the textual content of a post so that the visual output is contextually related to the accompanying professional content.

6.3 Semantic Content Analysis. OpenAI-assisted analysis provides structured communication indicators based on the post text. These indicators support content analytics and client communication frameworks.

6.4 Industry News & Topic Suggestions. The platform may suggest publicly available professional news and industry-related topics based on the user's professional context in order to assist users in drafting relevant professional posts. Suggested content is always editable and requires human review before publication.

7. Human Oversight

ScoringMy does not automatically publish AI-generated text or images. Users decide whether to use, edit or publish generated content. Semantic analysis results are assistive indicators and should be interpreted in the relevant communication context.

8. No Automated Employment Decision-Making

The semantic analysis functionality is not designed to make automated decisions with legal or similarly significant effects on employees or users, nor to perform automated employee-performance evaluations. It analyzes communication content against defined categories and communication criteria.

9. Storage and Model Training

AI-generated outputs may be stored when required by platform functionality or when saved by the user. PeopleXBrand does not use customer prompts, posts or generated outputs to train foundation models of its own. Processing by third-party AI providers is subject to the applicable commercial/API terms and data-protection arrangements for the services used.

10. Known Limitations

• generated text may contain factual inaccuracies or require editing
• generated images may not perfectly reflect the intended meaning of a post
• semantic classification may produce false positives, false negatives or context-dependent interpretations
• client-defined red lines and communication territories may require contextual human interpretation
• industry-news suggestions may be incomplete, outdated or require source verification.

11. Version History

Version	Date	Description
1.0	May 2026	Previous documented AI architecture
1.1	September 2026	Updated provider roles, data flows, semantic analytics and image generation

9.2. AI Risk & Safeguards Framework

Main risks associated with the use of AI and mitigation measures.

1. Purpose and Scope

This document describes the risk-management approach, safeguards and governance measures applied to ScoringMy's use of third-party AI services for text generation, image generation and semantic content analysis.

2. AI Providers and Functional Scope

• Anthropic Claude — text generation and adaptation.
• Google Gemini — image generation based on post text.
• OpenAI — semantic analysis of post text for communication analytics.

3. Data Protection Safeguards

AI requests are executed server-side. ScoringMy applies purpose limitation and data minimization. Claude receives only the professional and communication context needed for text generation; Gemini receives post text as context for image generation; OpenAI receives only the text of the post for semantic analysis, without associated employee identity metadata.

4. Human Oversight and User Control

Generated text and images are assistive outputs. They are not automatically published by ScoringMy. Users remain responsible for reviewing factual accuracy, appropriateness, tone and compliance with applicable internal communication policies before publication.

5. Semantic Analysis Safeguards

Semantic analysis can classify content, detect company references and assess content against client-defined communication red lines and territories. Because language is context-dependent, classifications may be imperfect. Results are treated as communication indicators rather than authoritative determinations.

• no associated employee identity metadata is sent to OpenAI for semantic analysis
• results should be interpreted in context and may require human review
• the functionality is not designed for automated employment decisions or employee-performance evaluation
• client-defined categories, red lines and territories should be sufficiently clear to reduce ambiguous classification.

6. Risk Scenarios and Mitigations

Risk scenario	Potential impact	Primary safeguard
Text hallucination / factual error	Misleading professional content	Human verification before publication
Image-content mismatch	Misleading or unsuitable visual	User review before use
Semantic misclassification	Incorrect category or territory	Contextual review; structured criteria
False red-line flag / missed flag	Incorrect compliance indicator	Treat result as assistive; human interpretation
Ambiguous communication territory	Inconsistent classification	Client-defined taxonomy and contextual review
Over-reliance on AI analytics	Analytics treated as definitive judgment	Transparency and non-decision-making limitation
News/source inaccuracy	Outdated or unreliable content	Source and factual verification before publication
Bias / prompt sensitivity	Non-neutral or variable output	Controlled prompts and user review

7. Model Training and Third-Party Processing

PeopleXBrand does not train or fine-tune foundation models using customer prompts, posts or generated outputs. Processing by third-party AI providers is governed by the commercial/API terms and data-protection arrangements applicable to the services used.

8. Compliance Position

ScoringMy is an AI-enabled application layer and not a provider of general-purpose AI models. Its features are designed around transparency, human oversight, data minimization and controlled professional use. Applicable regulatory obligations are assessed according to the actual intended use and deployment context.

9.3. AI Model Limitations Matrix

Known model limitations and the controls applied.

1. Purpose

This document describes known limitations of the third-party AI services integrated into ScoringMy and the mitigation measures applied at application level.

2. Limitations Overview

Category	Description	Potential impact	Mitigation
Text hallucination	Claude may generate plausible but inaccurate statements	Incorrect or misleading post	Human factual review
Context misinterpretation	Professional context or client red lines may be interpreted imperfectly	Irrelevant or misaligned draft	Structured prompts; user editing
Image mismatch	Gemini may create an image that only partially reflects post meaning	Unsuitable visual association	User review before use
Semantic misclassification	OpenAI may classify a post incorrectly	Wrong content category or territory	Treat as assistive; contextual review
Red-line false positive/negative	Semantic analysis may incorrectly flag or miss a communication restriction	Incorrect compliance indicator	Human interpretation; clear client criteria
Company-reference detection	Indirect references may be missed or over-detected	Incorrect company-mention metric	Aggregate review and contextual validation
Ambiguous territories	A post may reasonably fit multiple communication territories	Inconsistent categorization	Client taxonomy; allow contextual interpretation
Incomplete/current knowledge	Models or news suggestions may not reflect latest facts	Outdated or incomplete content	Verify sources and factual claims
Bias	Outputs may reflect model bias	Non-neutral suggestion or classification	Human review; controlled use
Prompt sensitivity	Small prompt differences may change outputs	Variability	Backend-controlled prompt templates
Over-reliance on AI	Users may treat generation or analytics as authoritative	Unverified publication or decisions	Transparency; human oversight

3. Provider Context

The limitations above relate to the capabilities used in ScoringMy: Anthropic Claude for text generation and adaptation, Google Gemini for image generation, and OpenAI for semantic analysis of post text.

4. Operational Safeguards

• human review of generated text and images before publication
• controlled backend prompt construction
• data minimization by functionality
• transparent treatment of semantic analytics as assistive indicators
• no automated publication
• no intended automated decision-making with legal or similarly significant effects on individuals.

9.4. AI Transparency Statement

Public explanation of how ScoringMy uses artificial intelligence.

1. Purpose

This statement explains in concise terms how artificial intelligence is used within ScoringMy, the role of each provider and the safeguards applied.

2. Role of Artificial Intelligence in ScoringMy

• drafting and adapting professional text content
• generating images related to professional post content
• suggesting content ideas and publicly available industry topics
• providing professional communication insights
• performing semantic analysis of post text for content classification and communication analytics.

3. AI Model Providers

• Anthropic Claude: text generation and adaptation.
• Google Gemini: image generation based on the textual content of a post.
• OpenAI: semantic analysis of post text.

4. What Semantic Analysis Does

OpenAI-assisted semantic analysis may classify posts as corporate, personal, educational or informational; identify whether content refers to the company; assess alignment with communication red lines defined by the client; and categorize posts within communication territories established by the client.

Only the text of the post is sent to OpenAI for this analysis. Associated employee name, email address, profile URL, author identity and other identifying metadata are not included in the semantic-analysis request.

5. Data Used for Content Generation

For Claude text generation, the platform may use professional and communication information optionally provided by the user, such as positioning goals, preferred topics and approaches, keywords, communication style, language and value proposition. Optional client-defined communication red lines may also be used to constrain generated content.

For Gemini image generation, the textual content of the post is used as context to generate a related image.

6. Human Oversight

AI-generated text and images are assistive outputs. ScoringMy does not automatically publish them. Users remain responsible for reviewing, editing where appropriate, verifying factual claims and deciding whether content should be published.

7. Analytics and Automated Decisions

Semantic-analysis outputs are communication indicators and may be imperfect. They are not designed to make automated decisions with legal or similarly significant effects on individuals and are not intended as automated employee-performance evaluations.

8. Data Protection and Responsible Use

ScoringMy applies data minimization and purpose limitation to AI requests. Users should not introduce sensitive personal data or unnecessary confidential information into AI-assisted features. AI interactions are performed through server-side API integrations.

9. Known Limitations

AI-generated content may contain inaccuracies, generated images may not perfectly reflect intended meaning, and semantic classifications may be context-dependent or incorrect. Users and clients should interpret AI-assisted outputs in context and verify relevant facts before external use.

10. Compliance Position

ScoringMy does not develop or provide general-purpose AI models. It integrates third-party AI services as an application layer and is designed around transparency, human oversight, data minimization and responsible professional use. Applicable legal obligations depend on the actual intended use and deployment context.

9.5. Responsible Use Guidelines for Brand Ambassadors

Guidelines for the responsible use of AI-assisted content and features.

1. Purpose

These guidelines apply to professionals participating in programs that use ScoringMy for digital presence, content creation and communication analytics.

2. AI-Assisted Features

ScoringMy may assist users with text generation through Anthropic Claude, image generation through Google Gemini, and professional content/topic suggestions. The platform may also use OpenAI to perform semantic analysis of post text for communication analytics.

3. Ambassador Responsibility and Human Review

Participants remain responsible for reviewing generated text and images, checking factual accuracy, ensuring alignment with organizational communication policies and deciding whether content should be published. ScoringMy does not automatically publish AI-generated content.

4. Recommended Use

• use AI drafts as a starting point rather than final authoritative content
• adapt tone and wording to the user's own professional voice
• verify sources, dates and factual claims when using industry news
• review generated images for relevance, accuracy and appropriateness
• respect client-defined communication policies, red lines and communication territories.

5. Prohibited / Inappropriate Use

• introducing confidential company information where it is not necessary
• introducing sensitive personal data or unnecessary personal data of third parties
• publishing defamatory, deceptive or knowingly false claims
• presenting unverified news or statistics as established fact
• using generated images in a misleading way or in a way that infringes third-party rights.

6. Data Protection

Users should not include sensitive personal data or unnecessary confidential information in AI-assisted interactions. For OpenAI semantic analysis, ScoringMy sends only the post text and does not include associated employee identity metadata in the analysis request.

7. Communication Analytics

Semantic classifications, company-reference indicators, red-line assessments and communication-territory classifications are assistive analytics. They may require contextual interpretation and should not be treated as definitive judgments about an employee or their performance.

8. Pre-Publication Checklist

• Is the text factually accurate and appropriately sourced?
• Does the post comply with company communication policies and red lines?
• Is the tone consistent with the ambassador's professional voice?
• Is the generated image relevant and non-misleading?
• Have confidential, sensitive or unnecessary personal data been excluded?
• If the post refers to news, have the source, date and context been verified?

9.6. Política de Uso Responsable para Embajadores

Spanish version of the responsible use guidelines, with the same content as section 9.5.

10. Platform security and privacy

AI features benefit from ScoringMy's general security controls, including authentication, access control, protected communications, monitoring and security procedures.

Full information on infrastructure, privacy and security measures is available on Security and Data Protection.

11. Contact

• PeopleXBrand Aceleradora S.L.
• C/ Gregorio Benítez 10, 28043 Madrid, Spain
• Tax ID B-56994916
• AI, Privacy & Compliance: legal @scoringmy. com