Principles for the responsible use of artificial intelligence on the AlexBodi.com website and its companies
1. Publisher, purpose and scope
This statement describes the principles of
SCANDIC ASSETS FZCO
Dubai Silicon Oasis DDP
Building A1/A2
342001 Dubai / UAE
Business License Number 050-3124863, for the responsible use of artificial intelligence in connection with https://www.AlexBodi.com and the own business processes recorded in each case.
Contact:
[email protected]
Telephone: +971 56 929 86 – 90.
It distinguishes legal requirements from voluntary, more extensive standards. It is neither an official approval nor a certificate of conformity for every conceivable AI system. Nor does it prove that a chatbot, an automated decision system or certain generative models are already used on the website. The original AI regulation 2024/1689 must be read together with the amending regulation 2026/1744. The latter was published on 24. July 2026 was published in the Official Journal and entered into force on 27. July 2026 in force. The following dates must therefore not be taken solely from older overviews.
Actual scope of application:
The specific information about a function used is provided where the processing or interaction takes place, in this case the responsible person is: SCANDIC DATA.
2. Human responsibility instead of shifting responsibility
AI should support people and should not serve as an excuse to abdicate responsibility for decisions. Persons responsible for the purpose, selection, configuration, release and monitoring of an application must be determined. A technically automatically generated output is not accurate, fair or lawful simply because of the way in which it is produced.
For decisions that have a significant impact on people, the necessary human intervention and review options must actually exist. Formal confirmation of an automatically generated recommendation without sufficient time, information and discretion is not meaningful human control.
Binding business declarations, sensitive specialist information and publications with significant reputational or damage potential are only made after an assessment appropriate to their risk. Every check by SCANDIC ASSETS FZCO evaluates the statement and the sources used and is not limited to linguistic plausibility.
3. Legal framework and timing
If the scope of application is relevant, the regulation (EU) 2024/1689 on artificial intelligence, in the version amended by the regulation (EU) 2026/1744, is particularly relevant. The specific obligations depend on the role, purpose, context of use, risk class and date of application. The General Data Protection Regulation, consumer protection, copyright, competition, product and, if applicable, other sectoral rules also apply.
A company headquarters outside the EU does not generally exclude the AI regulation. In particular, the provision of systems or models on the Union market and the legally covered use of results in the Union must be examined. Likewise, not every software automatically becomes a legally equivalent system through the use of the advertising word “AI”.
The fundamental bans on certain AI practices have been in effect since 2. February 2025; general transparency requirements according to article 50 basically since 2. August 2026. The amendment to 2026 provides, among other things, for special later dates for high-risk requirements and certain newly introduced bans. These transitional rules are not understood as permission to disregard existing data protection, criminal or personality laws.
4. Risk assessment before introduction and change
Before use, it must be clarified what specific benefits an application offers, which people are affected, which data is processed and what damage can result from errors or misuse. The risk assessment takes into account not only technical failures, but also discrimination, deception, loss of confidentiality and undue reliance on automated recommendations.
A system register should document the provider, version, purpose, data flows, affected groups of people, internal responsibility, risk assessment and release. Significant changes in purpose, new data sources, a change in vendor, and significant model updates require reconsideration. Merely releasing a product for one use case does not allow any other purposes.
Where a data protection impact assessment, fundamental rights impact assessment, conformity assessment, registration or other special assessment is required by law, it is not replaced by this ethics statement. Whether such an obligation exists must be determined on a case-by-case basis. A general “AI ethics seal” is not claimed.
5. Uses that are not permitted and excluded by us
We do not use AI for unlawful manipulation, exploitation of special vulnerability, unlawful discriminatory evaluation or other practices prohibited by law. In particular, legal prohibitions are not circumvented by the function in question being operated by an external provider or outside the EU.
As an internal company standard, we particularly exclude deceptive impersonation for fraud purposes, the creation of non-consensual intimate representations, depictions of child sexual abuse and unlawful invasions of privacy. This standard applies regardless of whether a specific supplementary AI rule is already applicable or whether its transition date is yet to come.
Biometric identification, emotion analysis, the derivation of specially protected characteristics and systems for evaluating people must not be introduced as casual convenience functions. They require particularly careful examination and are excluded if there is a legal ban. General website consent may disallow procedures prohibited by law.
6. Transparency in interactions and synthetic content
If a direct interaction with an AI system is recorded accordingly, it must be clear in good time that no natural person is responding, unless this is already obvious according to legal standards. Users should understand the role and essential limitations of the function and how to reach a responsible person.
We comply with the relevant labeling requirements for artificially created or manipulated images, audio files and videos. This particularly applies to content that constitutes deepfakes in the legal sense. A purely decorative or artistic use is not generally exempt from all transparency requirements; The adapted disclosure options provided for in each case must be specifically examined.
For AI-generated or manipulated texts intended to inform the public about matters of public interest, the special rules of Article 50 must be observed. A legal exception to human review or editorial control and responsible publication may not automatically be transferred to the separate rules for deepfakes.
Provider-side machine-readable labeling and human-recognizable disclosure by the user are different tasks. A technical marking does not always replace a required visible indication. Notices must be timely, clear and accessible according to the context of use.
7. Advertising, media and truthful representation
Generated visualizations may not be presented as photographic evidence of an actual advertising campaign, an encounter, the conclusion of a contract, a product test or a specific local situation if this actual event is not documented. An impression of authenticity is not proof of the reality depicted.
For example, clear references such as “AI-generated visualization” or “conceptual representation – not documentation of a real installation” are suitable for corresponding content, as long as they accurately describe the specific content. What remains crucial is a design that does not make the relevant overall impression misleading.
We do not use fictitious customer reviews or artificially generated recommendations masquerading as actual experiences. The use of names, voices and images requires a solid legal basis. AI-generated editing does not eliminate personal rights or rights to source materials.
The media surveillance by Prof. Dr. Stefan Haupt, Reinbeckstraße 21, D-12459 Berlin, Federal Republic of Germany, is carried out within the agreed scope of the order. Its mention does not constitute official recognition of the AI processes and does not constitute proof that each individual edition was legally examined before publication.
8. Data protection, confidentiality and data quality
Personal data will only be processed for specified purposes on a viable legal basis. Data minimization, transparency, accuracy, storage limitations and security must also be taken into account when using external AI services. The ability to find information publicly on the Internet does not mean that it can be freely processed for any AI purpose.
Special categories of data, information about criminal offenses and sensitive business documents may not be entered into AI systems without the necessary requirements. A general claim that sensitive data is needed to improve fairness is not enough. Special legal options for identifying and correcting distortions are also tied to strict requirements and protective measures.
Before a connection, provider roles, order processing, own provider purposes, data locations, subcontractors, deletion options and possible training use are checked. Processing abroad requires additional legal precautions, where relevant. The mere use of a European user account does not prove exclusively European data processing.
The inclusion of this statement on the website does not constitute consent to model training, profiling or automated decisions. Specific information and any necessary consents are separated and provided at the right time.
9. Fairness and protection against unjustified disadvantage
We check whether data, model behavior or operating conditions can unfairly disadvantage certain groups of people. Meaningful test cases, the actual usage context and the information limits of the available data are taken into account. A high average accuracy does not rule out significant disadvantages for individual groups.
Countermeasures should address the cause, for example through better data, restricting the intended use, changing decision thresholds or additional human testing. The lawful handling of data remains a prerequisite. An uncontrolled additional collection of protected characteristics is not a general license for fairness tests.
People who suspect incorrect or inappropriate treatment should be given a clear path to clarification. Rights to rectification and, if necessary, to human intervention must not be invalidated by reference to a supposedly unchangeable system result.
10. KYC, KYB and AML in collaboration with CRIF
The contractually agreed cooperation with CRIF Germany, https://www.crif.de, may concern procedures for customer identification (KYC - Know Your Customer), company auditing (KYB - Know Your Business) and money laundering prevention (AML - Anti-Money Laundering). Whether and which AI is actually used here must be determined based on the specific service. Collaboration alone does not substantiate automated decision-making about website visitors.
For personal results, sources, risks of confusion, timeliness and possible consequences must be checked. A sanction or identity hit may not be treated unchecked as established personal misconduct. For legally or similarly significant, exclusively automated decisions, the requirements and protective rights of Article 22 GDPR also apply, insofar as this is applicable.
Not every company audit is automatically a high-risk AI system. However, the assessment of the creditworthiness of natural persons may fall under the relevant categories of Annex III; The legal boundaries, for example for detecting financial fraud, must be examined carefully. The risk class is not derived solely from the name of the service provider or the term “compliance”.
11. Technical security with Darktrace
The technical monitoring of certain applications is carried out by Darktrace in accordance with the security architecture actually set up, https://www.darktrace.com, supports. AI-supported security functions can detect anomalies and facilitate technical analysis. Your results remain dependent on configuration, data quality, coverage and operational context.
Automatic protection measures must be appropriate to the risk. In the case of intervention-intensive reactions, false alarms, impairments of authorized users and human options for withdrawal or escalation must be taken into account. Safety data may not be repurposed for unrelated performance or behavioral profiles of employees without a separate review.
The use of Darktrace is not a guarantee of complete attack detection or certification of the legitimacy of all connected applications. Secure development, permissions, updates, emergency planning and traceable incident handling remain necessary.
12. Products, research and professional boundaries
The contractually agreed product testing with HP Prof. Dr. rer. nat. Dieter Lazik from the Slow Medicine Research Institute, Olympic Trail 7, 14471 Potsdam, refers to the products and test items actually commissioned. AI-generated summaries or hypotheses may not be presented as research results confirmed by this collaboration in the absence of supporting evidence.
AI expenditures are not a substitute for legally required testing procedures, reliable studies or professionally responsible product evaluations. In particular, an automatically formulated health benefit does not constitute a permissible medical or food advertising claim. By being named, an auditor does not automatically become a provider or conformity auditor of an AI system.
13. Competence, organization and traceability
We promote the AI competence of the people who use systems on our behalf that are tailored to tasks and risks. This particularly includes understanding boundaries, privacy, security, source verification, discriminatory errors and required human intervention. Suitable measures can include training, specific work instructions and supervised practical exercises.
Article 4 of the AI Regulation, as amended since 2026, requires measures to support the development of appropriate AI competence. This declaration does not promise a guaranteed individual level of knowledge for each person. Measures taken and open training needs should be documented in a comprehensible manner.
For risk-relevant applications, releases, essential tests, known limits, incidents and changes should be documented. The scope and storage periods depend on legal obligations and actual necessity; There is no provision for unlimited storage of all entries simply for reasons of proof.
14. Notes, review and improvement
References to problematic AI content or decisions can be made [email protected] with the subject “AI / Examination”. The affected function or release, time and a description of the problem are helpful. Please submit only necessary and lawfully available information.
Detected errors should be corrected appropriately and their causes investigated. If there are significant risks, a restriction of function, additional human testing or suspension of use may be considered. Legal reporting obligations and rights of data subjects remain independent of internal processing goals.
The principles will be reviewed in the event of relevant technical or legal changes. They do not constitute a claim that all organizational measures described have already been fully implemented or externally certified.
Status: 01. October 2026