OCEG’s Post

Twenty years ago, a new category of technology applications came into being to support GRC processes. AI adoption will do the same as it drives the need for a new ecosytem of solutions to manage, govern, manage risks, and ensure compliance. New tools are essential. Read my latest blog about Purpose-Built Solutions for governing AI. https://lnkd.in/eMc3274p

🎙 Excited to share that Holistic AI's Governance Solutions Included in the OECD.AI's Trustworthy AI Toolkit The catalogue provides one-stop-shop for helpful approaches, mechanisms and practices for trustworthy AI.    Our following tools are featured include: 👉 Holistic AI Governance Platform: a scalable solution to manage the risks of AI and empower trust  👉 Holistic AI Bias Audits: fully independent and impartial audits that ensure compliance with New York City Local Law 144 and other upcoming regulations.  👉 Holistic AI Audits: a bespoke AI risk audit solution comprising deep technical, quantitative analysis.  👉 Holistic AI risk mitigation roadmaps: guides to help enterprises mitigate some of the most common AI risks, presenting step-by-step solutions to protect against technical risks.  👉 Holistic AI Open-Source Library: an open-source tool to measure and mitigate bias across a variety of tasks.    Read the full release here 👇   #OECD #AIGovernance #AIRiskManagement #OpenSource #AISafety #TrustworthyAI  #ResponsibleAI

Compliance with OECD.ai principles can help businesses build trust, manage risks, stay ahead of regulations, gain competitive advantage, improve decision-making, and foster innovation and collaboration, all of which are critical in today’s rapidly evolving AI landscape. One more reason why Holistic AI is leading the pack for #aigovernance. I'm in Toronto and Boston April 15 - 19th and keen to meet with you to discuss how Holistic AI's Governance Platform supports innovation while controlling harm.

Scaling AI and ML initiatives requires robust governance to manage risk, drive compliance, and build organizational trust. Yet, existing frameworks often fall short in practical application, complicating rather than mitigating risks. With 95% of enterprises needing governance upgrades, automation emerges as a strategic enabler, streamlining access controls, policy enforcement, and auditability. By embedding automation into governance, organizations can elevate AI maturity, reduce risks, and accelerate the transformative impact of AI across their operations.#ResponsibleAI https://lnkd.in/gsu-FNW2

Calvin Risk Secures $4 Million as Its Mission to Make Enterprise AI Trustworthy Takes Off https://lnkd.in/dbAyGB9Y #AIinnovations #AIPoweredSecurity #TechSolutions

📢 Pleased to share that a number of our Holistic AI solutions are featured in the OECD.AI’s Catalogue of AI Tools and Metrics The catalogue provides one-stop-shop for helpful approaches, mechanisms and practices for trustworthy AI.    🚀 Our following tools are featured:  → Holistic AI Governance Platform: a scalable solution to manage the risks of AI and empower trust  → Holistic AI Bias Audits: fully independent and impartial audits that ensure compliance with  New York City Local Law 144 and other upcoming regulations.  → Holistic AI Audits: a bespoke AI risk audit solution comprising deep technical, quantitative analysis.  → Holistic AI risk mitigation roadmaps: guides to help enterprises mitigate some of the most common AI risks, presenting step-by-step solutions to protect against technical risks.  → Holistic AI Open-Source Library: an open-source tool to measure and mitigate bias across a variety of tasks.     Read the full release here 👉 https://lnkd.in/d5TVirEQ    #oecd #aigovernance #airiskmanagement #opensource #trustworthyai #ethicalai #responsibleai 

We are pleased to announce the release of our latest product, Ask ARIA Co-Pilot, a highly intuitive, accurate Conversational AI for first, second and third-line professionals. 4CRisk.ai uses language models trained on risk, compliance, and regulatory corpora with an informational retrieval component based on role-based access controls to create this safe, closed-domain, highly reliable Conversational AI tool. https://lnkd.in/e6xbSCCm https://lnkd.in/ee7WrU4Z #AI #regtech #risk #compliance #aipoweredcompliance #conversionalai #regulatorycompliance #regulatoryintelligence

Game-changer:ask the complex compliance / risk question and Aria researches your policies, standards, procedures, etc. to provide the most logical answer with full traceability to the source documents! In seconds. Your data, your models. Responsible and effective AI!

Will existing AI governance frameworks address the unique risks of agentic AI? 'Traditional' AI governance frameworks, such as the EU AI Act, NIST and ISO, were developed during a time when traditional machine learning was prevalent. Interestingly, there is no mention of the words 'agent' or 'agentic' in the EU AI Act, ISO 42001 or the NIST AI Risk Management Framework. They assume a context in which machine learning models are trained and developed to solve a specific problem. Once they are adequately tested and proven to work, they are deployed in production and monitored for performance issues. Although this traditional machine learning lifecycle remains relevant, it has been disrupted by generative AI. Now, much AI development involves leveraging pre-trained models, via APIs, and using techniques like fine-tuning and RAG to optimise these models for specific use cases or general purpose applications. Agentic AI will further disrupt how AI is developed and deployed, as well as the impact it can have. In a nutshell, agentic AI systems can autonomously develop plans, solve problems and execute tasks in a range of other applications which they are integrated with. For example, imagine an AI agent managing your calendar, scheduling meetings and sending messages. Although there is a lot of hype and the technology is immature, now is a good time to start thinking about the risks and challenges. Crucially, do our existing AI governance frameworks enable us to answer the following questions: ➡️ How can you have meaningful human oversight if agentic AI systems are autonomously executing tasks which can have a real-world impact? ➡️ Are there specific tasks or actions which agentic AI systems should not be allowed to perform? ➡️ Are there specific applications or data sources which agentic AI systems should not have access to? ➡️ Is it feasible to perform and continuously update traditional risk assessments for the countless multi-agentic AI systems which could become embedded across all software applications? ➡️ Should all employees have access to a personalised AI agent, trained on their data? I do not have the answers to all these questions. However, it is imperative for the AI governance community to take these head on. It is safe to assume that existing AI governance frameworks and approaches will require meaningful update, to address the risks and challenges of agentic AI.