Humaina’s Post

#2: How to Navigate the EU AI Act's Risk-Based Approach: A Practical Guide for Businesses When I first studied the EU AI Act's risk-based approach, it felt like deciphering a new language. Yet, this approach is the cornerstone of the Act, distinguishing between different AI applications based on their potential impact on rights and safety. 🔍 The Act categorizes AI systems from minimal to unacceptable risk, focusing regulatory efforts on high-risk applications. This ensures that AI's innovation potential is not stifled by overregulation while safeguarding public interests and fundamental rights. 🚀 For companies, understanding where your AI applications fall within this framework is crucial. High-risk categories include AI systems used in critical infrastructure, education, healthcare, employment, and law enforcement, among others. These systems require strict compliance with transparency, data governance, and accountability standards. 🛠️ Actionable Insight:  -              Conduct a thorough risk assessment of your AI systems. -              Focus on transparency, especially in how decisions are made.  -              For high-risk applications, ensure robust data management practices and prepare for rigorous testing and documentation requirements. 🇪🇺📘Resources EU AI Act:  Title II: Prohibited Artificial Intelligence Practices; Article 5 (Prohibited AI practices) Title III, Chapter 1: High-Risk AI Systems; Article 7 (High-risk AI systems) Title IV: Transparency Obligations for Certain AI Systems; Article 8 (AI systems posing limited risk); Article 9 (Minimal risk AI systems) 👍 Like to support a balanced approach to AI regulation, 💬 comment with your experiences in assessing AI risks. 💜 Follow my profile and stay tuned for the next post. #RiskBasedApproach #EUAIAct #AIEthics #ComplianceJourney #TechInnovation Disclaimer: This content is for informational purposes only and is not intended as legal advice. We make no representations as to accuracy or completeness. Consult a legal professional for advice on specific issues. Use of this information is at your own risk.

Thoughtful discussions on the topic “Data as new frontier of IP strategy in the age of AI” during the #IAM conference in London. Indeed data is the oil driving AI applications, and how to gather, execute and analyse on it is key elements. Some take aways around this from the sessions to highlight were: ⚠️The data used for training crucial and often more important than the LLM model itself. ⚠️ We need to legally find / acquire the right data from sources and providers. This includes models for how to contribute to the creator of generic data being trained on. ⚠️ If user behaviour data is to get used to improve modelling, it must be clearly explained to the user why you need data and how it will be used, processed and altered. ⚠️ On the corporate side important to establish a Text and Data Management policy. How to gather, process and maintain data for AI applications. ⚠️ Always consider; is generated results from an AI tool subject to IP rights or not. The more complex the answer, the more likely to be dependent on existing rights. ⚠️ Make results more transparent and indication on from data is derived Why not add source information, especially indication on of response is a hallucination. ⚠️ A domain that must evolve; How to make possible for users to withdraw a contribution, eg to a LLM model, and how to undo negative bias & errors. Many thanks to the panel elaborating on the topic; Yann Dietrich, Radomir Pivoda & Juan Becerra, all nicely moderated by Jason Raeburn. #AI #data #compliance #copyright

The marketplace value of a comprehensive AI solution like "Beyond-AI" for the legal sector is substantial, offering transformative capabilities in research, document management, and decision-making for legal professionals. Key Benefits and Impact: 1.      Enhanced Legal Research and Efficiency: Comprehensive Search Capabilities: ·      70 AI Search Services: Enables efficient access to diverse legal databases and sources. ·      AI Rankings: Prioritizes relevant and authoritative sources to save time and ensure quality. 2.      Advanced Language Models: ·      Multiple Chatbot LLMs: Offers varied perspectives on complex legal issues. ·      Paraphrasers and Summarizers: Simplifies the review process by rephrasing and summarizing complex texts. 3.      Improved Document Management: ·      Digital Document and Content Extraction: Automates document reviews and reduces errors, enhancing information organization and retrieval. ·      Grammar Correction: Ensures error-free, professional legal documents, reducing miscommunication risks. 4.      Ongoing Learning and Data Integration: Milvus Vector Database: ·      Continuous Learning: AI improves with usage, offering increasingly refined insights. ·      Storage and Retrieval: Quick access to historical data for better referencing and research building. 5.      Decision-Making and Productivity: ·      Collaborative Data Integration: Combines various data sources for comprehensive decision-making. ·      Downloadable and Storable Results: Archives research outcomes for future use, creating a valuable knowledge base. 6.      Sophisticated Summarization and Consolidation: ·      Quick access to synthesized information aids faster, informed decision-making. 7.      Competitive Advantage and Business Impact: ·      Efficiency and Cost-Effectiveness: Saves time and resources, increasing capacity to handle more cases and drive revenue growth. ·      Client Satisfaction and Retention: Enhances client satisfaction with faster, accurate services, showcasing a commitment to technology. ·      Cross-Sector Applicability: Beyond-AI’s adaptability extends its benefits to healthcare and other sectors, broadening its market appeal. Conclusion: "Beyond-AI" promises to revolutionize legal practices by improving research efficiency, document management, and decision-making. It enhances client satisfaction and firm profitability, modernizing legal workflows and providing a competitive edge in a tech-driven marketplace. https://lnkd.in/gBE2WAv9 Contact me at steven@corporate-payback.com or call 847-440-4439 for more information. #ChatGPT  #DataPrivacy #Healthcare #LegalRights #PublicHealth #MedicalRecords #HealthcarePolicy #HIPAA #FHIR #encryption #EMR #EHR #ccpa #gdpr #datacompliance  #cpra #grc #RegulatoryCompliance #DataPrivacy #CyberSecurity #DataProtection #ArtificialIntelligence #DigitalTransformation #CustomerRights

Why Combining Knowledge Graphs with RAG could be the Next Big Thing 🚀 RAG (Retrieval Augmented Generation) reduces GenAI hallucinations by grounding responses in data. It's a thing of beauty. Still, it often struggles with complex, multi-hop reasoning queries. What's a multi-hop reasoning query, you say? Well, here's an example: "Which departments need to adjust their procedures to remain compliant with the latest GDPR updates?" Unless there are particular texts that answer this query directly - highly unlikely - a RAG application will struggle to give a useful answer. Even a large LLM won't know the answer from memory, because the latest GDPR updates weren't part of its training data. Enter Knowledge Graphs (KGs.) KGs provide explicit knowledge representation in the form of a graph of relationships between entities. A KG for your business organization, say, would have departments, procedures, and regulations "nodes", linked together in a way that accurately and completely captures the relationships between these concepts. For example, that this procedure is used by that department. This enables sophisticated reasoning via the use of so-called 'reasoning engines' that use logical rules and ontologies to find answers. In other words - old school, cool AI. This particular example requires the AI to: 1. Identify multiple GDPR updates. 2. Map all the updates to current practices. 3. Determine departments connected to those practices. 4. Recommend adjustments. Traditional RAG struggles with such complexity, but some combination of KGs and LLMs/RAG may be the way forward. For example, an LLM could translate your query into a piece of code for searching the KG. Once that code has run, the same LLM could translate the resulting graph into natural language again, giving you a correct answer in plain English. It could even demonstrate how it arrived at the answer! I find the combination of LLMs and KGs promising and beautiful. One is general but inscrutable, the other is robust, yet incomplete. A potential match made in heaven. It might be a good idea to start scouting for tickets to the wedding. #AI #KnowledgeGraphs #RAG #Innovation #AIstrategy #DataDrivenGrowth

𝗧𝗵𝗲 𝗘𝗨 𝗔𝗜 𝗔𝗰𝘁: 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀, 𝗥𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗮𝘁𝗶𝗼𝗻𝘀, 𝗮𝗻𝗱 𝗖𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀 Redesigning of AI Systems and 𝗗𝗮𝘁𝗮 𝗚𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 by Louise de Leyritz The EU AI Act, due to come into force in 2024, will reshape AI systems and data management in the European Union - but not only there. . The article provides insights into the implications and focuses on the key challenges, recommendations and strategies for compliance. 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀: 𝗨𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 𝘁𝗵𝗲 𝗘𝗨 𝗔𝗜 𝗔𝗰𝘁: The legislation categorizes AI systems based on risk levels, posing challenges for organizations to navigate compliance. Recommendations include staying informed about the Act's provisions and assessing its impact on AI initiatives. 𝗜𝗱𝗲𝗻𝘁𝗶𝗳𝘆𝗶𝗻𝗴 𝗛𝗶𝗴𝗵-𝗥𝗶𝘀𝗸 𝗔𝗜 𝗦𝘆𝘀𝘁𝗲𝗺𝘀: Determining the high-risk status of AI systems is crucial for compliance. The Act employs a risk-based approach, outlining criteria for categorizing AI systems. Recommendations involve understanding the criteria and assessing AI systems accordingly. 𝗖𝗿𝗮𝗳𝘁𝗶𝗻𝗴 𝗮 𝗖𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 𝗔𝗰𝘁𝗶𝗼𝗻 𝗣𝗹𝗮𝗻: With the Act's rollout imminent, organizations must prepare compliance strategies. Data governance emerges as a cornerstone, with recommendations emphasizing data quality assurance, comprehensive documentation, and robust privacy measures. 𝗜𝗺𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗗𝗮𝘁𝗮 𝗚𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲: The EU AI Act underscores the significance of robust data governance practices in ensuring compliance. It mandates stringent requirements for data quality, data lineage, documentation, and privacy, necessitating proactive measures to align with regulatory mandates. 𝗞𝗲𝘆 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀: * Compliance with the EU AI Act is imperative for organizations deploying AI systems within the EU or impacting EU stakeholders. * Effective data governance, encompassing data quality assurance, comprehensive documentation, and robust privacy measures, is essential for compliance. * Proactive preparation and alignment with regulatory requirements are critical to navigate the implications of the EU AI Act successfully. 𝗖𝗼𝗻𝗰𝗹𝘂𝘀𝗶𝗼𝗻: As the EU AI law ushers in a new era of AI regulation, companies must prioritize data governance to ensure compliance and minimize risk. Through proactive measures and robust data governance practices, companies can navigate the complexities of AI regulation and realize the full potential of AI technologies. #AIRegulation #DataGovernance #Compliance #EUAIAct #DataQuality #Privacy #RiskManagement #AICompliance #TechnologyRegulation #DataManagement #DataEngineering

🤔 Because of the timing of its release, the interim report of the United Nations AI Advisory Body has received less attention. Its focus is on establishing guiding principles for global AI governance. 👉 Of particular interest is their Guiding Principle 3. AI governance should be built in step with data governance and the promotion of data commons. ✅Quote: “Data is critical for many major AI systems. Its governance and management in the public interest cannot be divorced from other components of AI governance ... Regulatory frameworks and techno-legal arrangements that protect privacy and security of personal data, consistent with applicable laws, while actively facilitating the use of such data will be a critical complement to AI governance arrangements, consistent with local or regional law. The development of public data commons should also be encouraged with particular attention to public data that is critical for helping solve societal challenges including climate change, public health, economic development, capacity building, and crisis response, for use by multiple stakeholders.” See https://lnkd.in/e6z2kGdp ➡️ Resonates a lot with the core recommendations of our essay (with Friederike Schüür)  on “Interwoven Realms: Data Governance as the Bedrock for AI Governance” See: https://lnkd.in/eZYgRKE2) 👉 Our essay provides six reasons why AI governance is unattainable without a comprehensive and robust framework of data governance. ➡️ In addressing this intersection, the essay aims to shed light on the necessity of integrating data governance more prominently into the conversation on AI, thereby fostering a more cohesive and effective approach to the governance of this transformative technology. 🤔I am eager to see how the AI Advisory Body will align AI Governance with Data Governance #ai #data #aigovernance #datagovernance #artificialintelligence

In a recent GovTech Review article, Hicksons Legal Digital Transformation Partner David Fischl and Solicitor Elias Dehsabzi discuss the rise of government agencies using Artificial Intelligence (AI) and Automated Decision Making (ADM) systems.   The post looks into the adoption of these technologies, key considerations and lessons to ensure agencies deliver efficient and fair public services in the digital age.   Read the article here: https://bit.ly/3TLW6vo   #AI #ArtificialIntellegence #AutomatedDecisionMaking #ChatGPT #Lawyer #Hicksons

🇪🇺 EU AI Act - What is it all about? In December, the European Union passed a pivotal regulation: the Artificial Intelligence Act. - and it will give the rhythm for a lot of companies this year. It's turning data governance from an "I'll do it later" task to a "let's get on this now" priority. As a data governance platform, we get a lot of questions: - What exactly does the EU AI Act entail? - Does it apply to my organization? - How will it influence data governance strategies? - What is the timeline for its implementation? - What are the compliance requirements? We've covered all these questions in the article below. This read by Louise de Leyritz should give you a well-rounded understanding of the matter. Enjoy the ride ⛷️

The document "Regulatory Sandboxes in Artificial Intelligence," published by the OECD in July 2023 was prepared by the #OECD Secretariat in consultation with the delegates of the Working Party on Artificial Intelligence Governance (#AIGO). 🔄 Source: https://lnkd.in/dm2xhjdk "#Sandbox" refers to a regulatory sandbox, a controlled environment where innovative products, services, or business models can be tested under a regulatory framework that is temporarily relaxed or modified to facilitate experimentation. ✅ Key Characteristics of the Sandbox: ➡ Temporary Nature: Sandboxes are established for a limited period, typically six months to two years, to allow sufficient time for testing and evaluation without prolonged exposure to reduced regulatory oversight; ➡ Participants operate within a controlled and supervised environment where certain regulatory requirements are waived or relaxed. This controlled setting ensures that potential risks are managed while allowing innovative solutions to be tested; ➡ Regulatory sandboxes involve close collaboration between regulators and participating firms; ➡ Participants receive specific regulatory relief tailored to their innovative products or services. This may include waivers from certain compliance requirements, exemptions from specific legal provisions, or modified regulatory obligations. ✅ For AI technologies, regulatory sandboxes are particularly useful because AI often intersects with multiple regulatory domains and involves complex ethical and technical considerations. AI sandboxes: ➡ Allow firms to test AI solutions in a real-world setting with oversight to ensure ethical use and compliance with relevant regulations. ➡ Enable regulators to develop a deeper understanding of AI technologies and their implications for privacy, security, and fairness. ➡ Facilitate interdisciplinary cooperation, involving data protection authorities, competition authorities, and sector-specific regulators, to comprehensively evaluate AI innovations. ✅ To effectively implement AI regulatory sandboxes, the document emphasizes several policy considerations: ➡Collaboration across different regulatory bodies and stakeholders is essential to address the multifaceted nature of AI. ➡Regulatory authorities need to build or enhance their technical expertise in AI to effectively manage sandbox activities. ➡Harmonizing sandbox frameworks across jurisdictions can prevent regulatory fragmentation and promote global standards. ➡Establishing clear and comprehensive criteria for sandbox eligibility and evaluation ensures that only genuinely innovative and beneficial projects are tested. #WhitePaper #Policies #Developers #Dev #AI #ArtificialIntelligence #Tips

AI’s potential to create positive human impact will depend on a responsible, human-centered approach that focuses on creating value for all. As it begins to transform industries, global policy makers are enacting legislation aimed at optimizing the opportunities while mitigating the risks of the technology. Find out more in our Artificial Intelligence global regulatory landscape report. #EY #ArtificialIntelligence #GenAI