[{"data":1,"prerenderedAt":700},["ShallowReactive",2],{"legal-ai-act-en":3},{"id":4,"title":5,"body":6,"description":692,"extension":693,"meta":694,"navigation":695,"path":696,"seo":697,"stem":698,"__hash__":699},"legal_en/en/legal/ai-act.md","Responsible AI & AI Act - VIKL",{"type":7,"value":8,"toc":656},"minimark",[9,14,22,29,35,38,42,49,54,67,74,79,90,94,101,106,126,133,142,144,148,152,155,159,162,169,173,184,191,193,197,201,208,228,232,235,241,261,268,272,275,282,284,288,292,299,303,323,327,335,339,342,355,357,361,368,372,402,409,411,415,421,496,503,505,509,513,519,522,526,541,545,548,566,573,575,579,648,650],[10,11,13],"h2",{"id":12},"our-commitment","Our commitment",[15,16,17,18],"p",{},"VIKL is an AI companion designed to support managers in everyday relational situations. We build our technology with a strong conviction: ",[19,20,21],"strong",{},"AI should serve human clarity, never replace it.",[15,23,24,25,28],{},"The European Artificial Intelligence Act (EU 2024/1689), known as the ",[19,26,27],{},"AI Act",", entered into force on August 1, 2024. It establishes a harmonized framework for the development and use of AI systems in the European Union. VIKL fully embraces this framework.",[15,30,31,34],{},[19,32,33],{},"Last updated:"," April 2026",[36,37],"hr",{},[10,39,41],{"id":40},"_1-vikls-risk-classification","1. VIKL's risk classification",[15,43,44,45,48],{},"The AI Act classifies AI systems across four risk levels: unacceptable, high, limited, and minimal. Based on our analysis, VIKL falls within the category of ",[19,46,47],{},"limited-risk AI systems",".",[15,50,51],{},[19,52,53],{},"Why?",[55,56,57,61,64],"ul",{},[58,59,60],"li",{},"VIKL does not make automated decisions with legal or significant effects on individuals.",[58,62,63],{},"VIKL is not designed for recruitment, performance evaluation, social scoring, biometric surveillance, or any of the use cases listed in Annex III of the regulation.",[58,65,66],{},"VIKL is a reflection and preparation tool: it generates suggestions, reformulations, and action plans that the manager freely chooses to adopt or not.",[15,68,69,70,73],{},"As such, VIKL is subject to the ",[19,71,72],{},"transparency obligations"," applicable to limited-risk systems (Article 50), and we are committed to complying with all general provisions of the regulation.",[75,76,78],"h3",{"id":77},"our-role-under-the-ai-act","Our role under the AI Act",[15,80,81,82,85,86,89],{},"Under the regulation, ",[19,83,84],{},"MINDGUARD PROJECT"," is the ",[19,87,88],{},"provider"," of the VIKL AI system — the entity that developed the system and places it on the market under its own name. VIKL relies on large language models provided by Azure OpenAI Service (GPAI models within the meaning of Articles 51-56), for which Microsoft/OpenAI are the providers. MINDGUARD PROJECT is not the provider of these underlying models.",[75,91,93],{"id":92},"intended-use-and-excluded-uses","Intended use and excluded uses",[15,95,96,97,100],{},"VIKL is designed for a specific purpose: ",[19,98,99],{},"supporting managers in preparing for and handling relational situations"," (team tensions, feedback, difficult conversations, performance discussions).",[15,102,103],{},[19,104,105],{},"VIKL is not intended and must not be used for:",[55,107,108,111,114,117,120,123],{},[58,109,110],{},"recruitment or candidate screening;",[58,112,113],{},"formal performance evaluation or employee scoring;",[58,115,116],{},"disciplinary, promotion, or termination decisions;",[58,118,119],{},"psychological or medical assessment;",[58,121,122],{},"employee surveillance or monitoring;",[58,124,125],{},"any decision with legal or significant effects on an individual.",[15,127,128,129,132],{},"Deliverables generated by VIKL (messages, action plans, scenarios) are ",[19,130,131],{},"thinking tools",". They should not be used as the sole basis for high-impact managerial decisions.",[134,135,136],"blockquote",{},[15,137,138,141],{},[19,139,140],{},"Reference:"," Articles 6, 7 and Annex III of Regulation (EU) 2024/1689",[36,143],{},[10,145,147],{"id":146},"_2-transparency","2. Transparency",[75,149,151],{"id":150},"you-know-youre-interacting-with-ai","You know you're interacting with AI",[15,153,154],{},"From the very first exchange, VIKL clearly identifies its nature: it is a companion powered by artificial intelligence. There is no ambiguity that responses are generated by an AI system.",[75,156,158],{"id":157},"how-it-works","How it works",[15,160,161],{},"VIKL uses large language models (LLMs) hosted on Azure OpenAI Service in Europe. These models analyze the context you describe and generate tailored suggestions, drawing on methodologies from organizational psychology and mediation.",[15,163,164,165,168],{},"VIKL does not make decisions for you. Every suggestion, reformulation, or action plan is a ",[19,166,167],{},"thinking tool"," that you remain free to use, adapt, or disregard.",[75,170,172],{"id":171},"system-limitations","System limitations",[55,174,175,178,181],{},[58,176,177],{},"VIKL does not have access to your company's actual context beyond what you share.",[58,179,180],{},"Suggestions do not constitute legal advice, medical advice, or HR prescriptions.",[58,182,183],{},"Like any LLM-based system, VIKL may occasionally produce imprecise or unsuitable responses.",[134,185,186],{},[15,187,188,190],{},[19,189,140],{}," Article 50 of Regulation (EU) 2024/1689",[36,192],{},[10,194,196],{"id":195},"_3-human-oversight","3. Human oversight",[75,198,200],{"id":199},"the-manager-remains-the-decision-maker","The manager remains the decision-maker",[15,202,203,204,207],{},"VIKL is designed as a ",[19,205,206],{},"decision-support tool",", not an autonomous system. The product's architecture has been built to maintain human oversight at every step:",[55,209,210,216,222],{},[58,211,212,215],{},[19,213,214],{},"No automatic actions",": VIKL does not contact anyone, send any messages, or modify any HR systems.",[58,217,218,221],{},[19,219,220],{},"Deliverables to review",": feedback messages, action plans, and generated scenarios are drafts that the manager reviews and adapts before use.",[58,223,224,227],{},[19,225,226],{},"No closed decision loop",": VIKL suggests, the manager decides and acts.",[75,229,231],{"id":230},"no-emotion-inference","No emotion inference",[15,233,234],{},"The AI Act prohibits the use of emotion recognition systems in the workplace (Article 5.1.f). This point is particularly relevant for VIKL, which supports managers in emotionally charged situations (tensions, conflicts, difficult feedback).",[15,236,237,240],{},[19,238,239],{},"VIKL does not infer its users' emotions."," Specifically:",[55,242,243,249,255],{},[58,244,245,248],{},[19,246,247],{},"No emotion detection",": VIKL does not analyze user text to deduce an emotional state (stress, anger, anxiety, etc.). No sentiment analysis or emotion recognition algorithm is applied to user messages.",[58,250,251,254],{},[19,252,253],{},"No emotional state classification",": VIKL does not produce any label, score, or emotional profile of the user — neither for the user, nor for a third party, nor for an employer.",[58,256,257,260],{},[19,258,259],{},"Technical safeguards",": specific measures have been built into the system architecture (framing instructions, output filters) to prevent any emotion inference behavior, including incidental inference.",[15,262,263,264,267],{},"VIKL helps managers ",[19,265,266],{},"describe and structure"," a relational situation. It is the user who qualifies the context, the stakes, and the dynamics at play. VIKL never attempts to guess what the user is feeling.",[75,269,271],{"id":270},"no-profiling-or-scoring","No profiling or scoring",[15,273,274],{},"VIKL does not rate users' managerial skills. Self-assessments (DISC, leadership, feedback) are voluntarily completed by the user to personalize recommendations, and are never shared with the employer.",[134,276,277],{},[15,278,279,281],{},[19,280,140],{}," Articles 5.1.f and 14 of Regulation (EU) 2024/1689",[36,283],{},[10,285,287],{"id":286},"_4-data-privacy-and-robustness","4. Data, privacy, and robustness",[75,289,291],{"id":290},"your-data-is-not-used-to-train-ai","Your data is not used to train AI",[15,293,294,295,298],{},"Conversations with VIKL are ",[19,296,297],{},"never used to train or fine-tune AI models",", neither by VIKL nor by our model providers. This guarantee is contractually established with Azure OpenAI Service.",[75,300,302],{"id":301},"european-hosting-and-encryption","European hosting and encryption",[55,304,305,311,317],{},[58,306,307,310],{},[19,308,309],{},"Hosting",": Azure Europe (France and Western Europe).",[58,312,313,316],{},[19,314,315],{},"Encryption",": data is encrypted in transit (TLS 1.2+) and at rest (AES-256). Conversations are protected so that neither the employer nor VIKL staff can access their content in clear text.",[58,318,319,322],{},[19,320,321],{},"Data isolation",": each organization's data is logically separated.",[75,324,326],{"id":325},"gdpr-compliance","GDPR compliance",[15,328,329,330,48],{},"VIKL complies with the General Data Protection Regulation (GDPR). For details on processing activities, legal bases, retention periods, and data subject rights, please refer to our ",[331,332,334],"a",{"href":333},"/en/privacy","Privacy Policy",[75,336,338],{"id":337},"data-quality-and-relevance","Data quality and relevance",[15,340,341],{},"The training data for the underlying models is managed by Azure OpenAI Service according to their own quality and governance procedures. VIKL enriches these models with structured methodologies (organizational psychology, mediation, leadership) to ensure response relevance within its domain.",[134,343,344,350],{},[15,345,346,349],{},[19,347,348],{},"Note:"," Articles 10 and 15 of the AI Act primarily apply to high-risk AI systems. Although VIKL falls under limited risk, we voluntarily draw on these requirements regarding data quality and robustness as best practices.",[15,351,352,354],{},[19,353,140],{}," Regulation (EU) 2016/679 (GDPR); Articles 10, 15 of Regulation (EU) 2024/1689 (adopted as best practices)",[36,356],{},[10,358,360],{"id":359},"_5-ai-literacy","5. AI literacy",[15,362,363,364,367],{},"Since February 2, 2025, Article 4 of the AI Act requires providers and deployers of AI systems to promote ",[19,365,366],{},"AI literacy"," among their staff and users.",[75,369,371],{"id":370},"how-vikl-contributes","How VIKL contributes",[55,373,374,380,386],{},[58,375,376,379],{},[19,377,378],{},"Explanatory onboarding",": every new user is informed about VIKL's AI nature, capabilities, and limitations.",[58,381,382,385],{},[19,383,384],{},"Ongoing transparency",": VIKL regularly reminds users that it is a thinking support tool, not a substitute for human judgment.",[58,387,388,391,392,396,397,401],{},[19,389,390],{},"Educational resources",": our blog regularly publishes content on responsible AI use in management (",[331,393,395],{"href":394},"/en/blog/ia-coaching-management","Can AI really help managers?",", ",[331,398,400],{"href":399},"/en/blog/coaching-individuel-vs-coaching-ia","Individual coaching vs AI coaching",").",[134,403,404],{},[15,405,406,408],{},[19,407,140],{}," Article 4 of Regulation (EU) 2024/1689",[36,410],{},[10,412,414],{"id":413},"_6-prohibited-practices","6. Prohibited practices",[15,416,417,418],{},"Article 5 of the AI Act prohibits certain AI practices deemed unacceptable. ",[19,419,420],{},"VIKL does not use any of these practices:",[422,423,424,437],"table",{},[425,426,427],"thead",{},[428,429,430,434],"tr",{},[431,432,433],"th",{},"Prohibited practice",[431,435,436],{},"VIKL's position",[438,439,440,449,457,465,473,481,489],"tbody",{},[428,441,442,446],{},[443,444,445],"td",{},"Subliminal or deceptive manipulation techniques",[443,447,448],{},"VIKL does not manipulate user behavior. Suggestions are explicit and modifiable.",[428,450,451,454],{},[443,452,453],{},"Exploitation of vulnerabilities (age, disability, social situation)",[443,455,456],{},"VIKL does not target or exploit any vulnerabilities.",[428,458,459,462],{},[443,460,461],{},"Social scoring",[443,463,464],{},"VIKL does not rate individuals and does not produce any behavioral score.",[428,466,467,470],{},[443,468,469],{},"Criminal risk assessment (predictive profiling)",[443,471,472],{},"Not applicable.",[428,474,475,478],{},[443,476,477],{},"Facial recognition and biometrics",[443,479,480],{},"VIKL does not process any biometric data.",[428,482,483,486],{},[443,484,485],{},"Emotion inference in the workplace (Art. 5.1.f)",[443,487,488],{},"VIKL does not infer its users' emotions. No sentiment analysis, emotion recognition, or emotional state classification is performed — neither from text nor from biometric data. Technical safeguards are built into the system to prevent any emotion inference behavior.",[428,490,491,494],{},[443,492,493],{},"Biometric categorization",[443,495,472],{},[134,497,498],{},[15,499,500,502],{},[19,501,140],{}," Article 5 of Regulation (EU) 2024/1689",[36,504],{},[10,506,508],{"id":507},"_7-governance-and-contact","7. Governance and contact",[75,510,512],{"id":511},"accountability","Accountability",[15,514,515,516,518],{},"VIKL is developed and operated by ",[19,517,84],{}," (SIRET: 988 560 009 00017), a French company based in La Garenne-Colombes (92), France.",[15,520,521],{},"AI Act compliance is overseen by the company's executive management, in coordination with our product and engineering teams.",[75,523,525],{"id":524},"regulatory-monitoring","Regulatory monitoring",[15,527,528,529,532,533,536,537,540],{},"We actively follow the work of the ",[19,530,531],{},"European AI Office",", the guidelines of the ",[19,534,535],{},"European Artificial Intelligence Board",", and publications from the ",[19,538,539],{},"CNIL"," (French data protection authority) to adapt our practices as the regulation is implemented.",[75,542,544],{"id":543},"contact-us","Contact us",[15,546,547],{},"For any questions regarding VIKL's AI Act compliance:",[55,549,550,560],{},[58,551,552,555,556],{},[19,553,554],{},"Email:"," ",[331,557,559],{"href":558},"mailto:privacy@vikl.ai","privacy@vikl.ai",[58,561,562,565],{},[19,563,564],{},"Address:"," MINDGUARD PROJECT, 73 Avenue Foch, 92250 La Garenne-Colombes, France",[134,567,568],{},[15,569,570,572],{},[19,571,140],{}," Article 26 of Regulation (EU) 2024/1689",[36,574],{},[10,576,578],{"id":577},"compliance-timeline","Compliance timeline",[422,580,581,594],{},[425,582,583],{},[428,584,585,588,591],{},[431,586,587],{},"Deadline",[431,589,590],{},"Obligation",[431,592,593],{},"VIKL status",[438,595,596,607,616,627,637],{},[428,597,598,601,604],{},[443,599,600],{},"February 2, 2025",[443,602,603],{},"Prohibition of unacceptable practices (Art. 5)",[443,605,606],{},"✅ Measures in place",[428,608,609,611,614],{},[443,610,600],{},[443,612,613],{},"AI literacy (Art. 4)",[443,615,606],{},[428,617,618,621,624],{},[443,619,620],{},"August 2, 2025",[443,622,623],{},"GPAI model provider obligations (Art. 51-56)",[443,625,626],{},"Not applicable (MINDGUARD PROJECT is the provider of the VIKL system, not of the underlying GPAI models)",[428,628,629,632,635],{},[443,630,631],{},"August 2, 2026",[443,633,634],{},"Transparency obligations (Art. 50)",[443,636,606],{},[428,638,639,642,645],{},[443,640,641],{},"August 2, 2027",[443,643,644],{},"High-risk system obligations (Art. 6-49)",[443,646,647],{},"Not applicable based on our analysis (limited risk)",[36,649],{},[15,651,652],{},[653,654,655],"em",{},"This document is provided for informational purposes and does not constitute legal advice. It will be updated as the regulatory framework evolves.",{"title":657,"searchDepth":658,"depth":658,"links":659},"",2,[660,661,666,671,676,682,685,686,691],{"id":12,"depth":658,"text":13},{"id":40,"depth":658,"text":41,"children":662},[663,665],{"id":77,"depth":664,"text":78},3,{"id":92,"depth":664,"text":93},{"id":146,"depth":658,"text":147,"children":667},[668,669,670],{"id":150,"depth":664,"text":151},{"id":157,"depth":664,"text":158},{"id":171,"depth":664,"text":172},{"id":195,"depth":658,"text":196,"children":672},[673,674,675],{"id":199,"depth":664,"text":200},{"id":230,"depth":664,"text":231},{"id":270,"depth":664,"text":271},{"id":286,"depth":658,"text":287,"children":677},[678,679,680,681],{"id":290,"depth":664,"text":291},{"id":301,"depth":664,"text":302},{"id":325,"depth":664,"text":326},{"id":337,"depth":664,"text":338},{"id":359,"depth":658,"text":360,"children":683},[684],{"id":370,"depth":664,"text":371},{"id":413,"depth":658,"text":414},{"id":507,"depth":658,"text":508,"children":687},[688,689,690],{"id":511,"depth":664,"text":512},{"id":524,"depth":664,"text":525},{"id":543,"depth":664,"text":544},{"id":577,"depth":658,"text":578},"Learn how VIKL complies with the European Artificial Intelligence Act (AI Act) and our commitment to responsible AI.","md",{},true,"/en/legal/ai-act",{"title":5,"description":692},"en/legal/ai-act","N_Z3AnGbrRrT8y-pzWPq0e0ctcezQAT-hHbCcRdXz2k",1775931406236]