CertifAI

𝐃𝐨 𝐲𝐨𝐮 𝐤𝐧𝐨𝐰 𝐭𝐡𝐞 𝐬𝐭𝐚𝐧𝐝𝐚𝐫𝐝 𝐟𝐨𝐫 𝐭𝐫𝐮𝐬𝐭𝐰𝐨𝐫𝐭𝐡𝐲 𝐚𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐢𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞? 💡 Trust, protection of health, safety and fundamental rights of EU citizens are the ambitious demands of the European AI Act on AI technology. For providers, deployers, manufacturers, importers and distributors of AI systems covered by the AI Act, the question is how to technically fulfill the requirements of the law in order to ensure the trustworthiness of their systems. Technical standards help to answer this question. Here is what you need to know about the current stage of technical AI standards: ➡️ Standards currently considered by the CEN-CENELEC Joint Technical Committee (JTC) 21 on artificial intelligence (AI) already cover most of the AI Acts requirements for trustworthy AI. This was shown by an Analysis of the preliminary AI standardization work plan in support of the AI Act by the Joint Research Center of the European Commission. Recommendations by experts aim to assist the closing of still open gaps by planned European standardization efforts. ➡️ The JTC 1/SC 42 of ISO and IEC is another one of the most relevant standardization bodies when it comes to AI. Based on inputs of a broad range of stakeholders on the desired societal and ethical outcomes of trustworthy AI, the committee’s horizontal technical standards focus on transparency, data quality, and reliability of AI systems. In a White Paper about AI across industries, the AI project team in the IEC Market Strategy Board, the German Research Centre for Artificial Intelligence and Haier Group provided specific recommendations targeted at the IEC with regard to AI standardization. Among others, the IEC is encouraged to coordinate the standardization of data semantics and ontologies, develop an AI reference architecture with consistent interfaces, and explore the potential for AI conformity assessment needs. ➡️ The example of the AI Act shows that technical standards will play a fundamental role in the integration of trustworthy and legally compliant AI into society. 🦾 If you are developing or using AI, we at CertifAI are looking forward to support your organizations in implementing the upcoming technical standards in your AI systems! #AI #AIRegulation #AIAct #AItrustworthiness #technicalstandard #standardization

🇪🇺 𝐇𝐨𝐭 𝐨𝐟𝐟 𝐭𝐡𝐞 𝐏𝐫𝐞𝐬𝐬: 𝐄𝐔 𝐀𝐈 𝐀𝐜𝐭 𝐏𝐮𝐛𝐥𝐢𝐬𝐡𝐞𝐝 𝐢𝐧 𝐭𝐡𝐞 𝐎𝐟𝐟𝐢𝐜𝐢𝐚𝐥 𝐉𝐨𝐮𝐫𝐧𝐚𝐥   More than 3 years after the first proposal, the EU AI Act has just been published today in the Official Journal of the European Union as Regulation (EU) 2024/1689. It will enter into force on 1 August 2024 and it will be applicable in parts already by February 2025.    𝐖𝐡𝐚𝐭 𝐝𝐨𝐞𝐬 𝐭𝐡𝐚𝐭 𝐦𝐞𝐚𝐧 𝐟𝐨𝐫 𝐲𝐨𝐮𝐫 𝐜𝐨𝐦𝐩𝐚𝐧𝐲 𝐧𝐨𝐰? 1️⃣ As an AI system provider or deployer, you need to ensure by February 2025 a sufficient level of AI literacy of your staff and other persons operating and using your AI systems. 2️⃣ Certain AI systems are about to be prohibited by February 2025. To ensure that you do not have prohibited AI system in operation in your company, a complete AI inventory should be created. 3️⃣ Other obligations will be applicable later in 2025 or in 2026 (e.g., extensive requirements for high-risk AI systems), which might take longer for compliance measures to be identified and implemented.   At CertifAI, we can support you with all of the above on your EU AI Act compliance journey. We are offering AI literacy workshops for your staff, helping you to create an exhaustive AI inventory and preparing you early for further EU AI Act obligations emerging down the line.   🔗 Official Journal: https://lnkd.in/e6VkKySk   #EUAIAct #AIAct #AI #ArtificialIntelligence #AIRegulation #TrustworthyAI

Everybody is talking about fairness requirements under the AI Act but how can you technically actually check your AI system on unrightful biases? Our CEO Robert Kilian and Lead Data Scientist Nico Schmidt had the honor to talk about "Testing Fairness Requirements According to the EU AI Act" during the workshop series "European AI Regulation and Its Operationalization," hosted by the DIN Deutsches Institut für Normung e. V..   From hiring processes to loan approvals, healthcare decisions, or the everyday use of an AI chatbot, AI systems are making decisions that have profound impacts on individuals and communities. But ensuring these systems are fair is not just a technical challenge. It's a crucial part of the long-term success of AI systems. From trust in your system, transparency, and the company’s reputation to legal and ethical compliance, fairness is an underlying dimension of every AI system.   You can find the whole presentation in the comments. Any questions? Feel free to contact us. With many years of experience and diverse expertise across different industries, CertifAI can help you establish fairness in your AI applications. #AIAct #KIVerordnung #AItesting #AIQualityAssurance

🙌🏼 𝐄𝐱𝐞𝐜𝐮𝐭𝐢𝐯𝐞 𝐄𝐝𝐮𝐜𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐜𝐮𝐬𝐞𝐬 𝐨𝐧 𝐀𝐈: 𝐀𝐈 𝐒𝐭𝐚𝐫𝐭𝐮𝐩 𝐏𝐢𝐭𝐜𝐡 𝐚𝐭 𝐄𝐒𝐌𝐓 𝐁𝐞𝐫𝐥𝐢𝐧   We are happy that our CTO, Jan Zawadzki, had the opportunity to pitch our solution for reliable AI engineering and AI testing at a specialised executive education program at ESMT Berlin.   Together with fellow AI startups, we participated in an evening of engaging discussions and learning, showcasing the transformative potential of AI to a group of diverse and visionary leaders from a leading automotive company.   The experience was not only about the potential of AI to lead with a long-term view, but also about inspiring collaborations in the future. We shared our startup stories and explored the endless possibilities AI will bring, if it's reliable.    A heartfelt thank you to ESMT Berlin, and particularly to Lei Lu, for the invitation. Connecting with industry pioneers who are focusing on integrating AI into their strategies was truly great. 🤝 #ai #aitesting #transformation #automotive #airegulation

🔩 𝐓𝐀𝐂𝐇𝐄𝐋𝐄𝐒: 𝐀𝐈'𝐬 𝐋𝐨𝐨𝐬𝐞 𝐒𝐜𝐫𝐞𝐰 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞 At the recent TACHELES conference (June 18, 2024) on autonomous and connected driving, our CTO, Jan Zawadzki, offered a slightly different perspective on AI, adding to the insightful and engaging contributions of esteemed speakers Sascha Pallenberg 潘賞世 and Axel Gern, whose remarks fueled a dynamic and thought-provoking debate. Jan introduced an analogy comparing AI to “loose screws” in a product; eventually, these loose screws will fail, potentially causing the product to break. This analogy resonated strongly with the audience, highlighting the inherent risks associated with integrating AI technology into a product. He emphasized that there are two options:   1️⃣ Only use the loose screw in a safe environment (e.g. trunk), where its failure won’t cause significant harm. 2️⃣ Attempt to tighten the loose screw, ensuring it remains secure and functional in order to use it in more risky environments.   At CertifAI, we specialize in the latter. Our mission is to help you securely tightening the “loose screws” in your products, ensuring they are reliable and market-ready. By leveraging our expertise, we provide the tools and knowledge needed to reinforce the integrity of AI components. Let’s tighten those loose screws together! 🛠️🔧 Photos: © FZI Forschungszentrum Informatik, Photographer Paul Gärtner

*Webinar on technical testing of fairness requirements under the AI Act and sectoral legislation with our data sciene lead Nico Schmidt and our CEO Robert Kilian AI models are taking on more and more tasks in personnel selection, credit risk, medicine and many other fields where discriminatory decisions are not only morally questionable, but can also have legal consequences. Possible unequal treatment in model outputs resulting from biases in the training data or errors in the models thus make fairness considerations a key pillar of the trustworthiness of current and future models. The workshop to be held tomorrow with KI.NRW and DIN Deutsches Institut für Normung e. V. 9.30am CET (IN GERMAN) will show how the fairness requirements from current law and the upcoming AI regulation can be implemented technically. Please register under the link below. #AItesting #AIfairness #AIAct https://lnkd.in/dm5iDrqq

AI and computer vision are revolutionizing how manual manufacturing companies gather shop floor insights. But what does the EU’s AI Act⚖️ mean for products like Deltia.ai? We’ve teamed up with Robert Kilian and the experts over at CertifAI to explore how AI regulatory compliance doesn’t have to be a pain – in fact, it can be an advantage. Read it here➡️ https://lnkd.in/dx_paFnb #AIAct #aicertification #airegulation #manualmanufacturing #ComputerVision #industry40 #AI

Bringing reliable and trustworthy AI solutions to manufacturing processes: CertifAI is working with Deltia on testing their AI-based products. AI and computer vision are revolutionizing how manual manufacturing companies gather shop floor insights. But what does the EU AI Act mean for products like Deltia in terms of software testing and how can AI testing actually help build an competitive advantage? Read Deltias blogpost on our work for this great client here ➡️ https://lnkd.in/dx_paFnb Dr. Silviu Homoceanu Robert Kilian Jan Zawadzki Stephanie Jonkers Nico Schmidt #AIAct #AItesting #aicertification #airegulation #manualmanufacturing #ComputerVision

WEBINAR ALERT: TESTING & ZERTIFIZIERUNG VON HOCHRISIKO-KI-SYSTEMEN am 17. Juni 2024 von 17.00-18.00 Uhr (Anmeldung unter dem Link in den Kommentaren) mit Jan Zawadzki und Robert Kilian   Das Testen von Hochrisiko-KI-Systemen ist der Schlüssel zur Identifizierung geeigneter Risikomanagementmaßnahmen und zur Einhaltung der Testing-Anforderungen aus dem AI Act und sektoralen Bestimmungen. In diesem Webinar werden wir folgende Fragestellungen behandeln und auch wieder eine Q&A anbieten: - Warum sollte ich mein KI-System testen lassen? Wann handelt es sich um ein Hochrisiko-KI-System und wie teste ich KI-Systeme? - Was sind mögliche Maßnahmen zur Risikominimierung und zur Qualitätssicherung? - Welche Mehrwerte bringt das Testing durch einen externen Partner? - Welche sektoralen Regularien sind zu berücksichtigen? Welche Industrien sind betroffen?

CertifAI CEO Robert Kilian at the German parliament Here is a brief summary of the opening statement mentioning important topics with regard to the testing and certification of AI systems: - Technical standardization is needed now. The harmonized standards should be ready by summer/autumn 2025 at the latest so that companies still have enough time to implement these standards before the regulatory transition periods end. This is a task of the industry organized in the standardization bodies but politics and administration should influence the industry to get more involved now! - Germany should work to ensure that internationally applicable ISO/IEC standards, in particular ISO/IEC 42001 and ISO/IEC 42006, become the basis for the AI Act harmonized standards. This is the only way to avoid fragmentation of the global market and serious disadvantages for European and German providers, particularly when exporting outside EU. - A number of national initiatives are already carrying out decisive preparatory work for technical standardization under the AI regulation. For example, MISSION KI working on high risk system criteria. - Germany is among the world's leading nations in the research area of AI safety and the underlying regulatory and technical research. But we need significantly more funding capacity now in the scientific and startup field to address the current issues. #aitesting #certifai