Dynamo AI

📝Dynamo AI's Research Presentation at NeurIPS 2025🧪 Ever wonder why LLMs perform so well on theoretical benchmarks but miss the mark in real-world use cases? Dynamo's Lead Research Engineer Blazej Manczak and ML Research team recently presented a paper at NeurIPS that exposes this gap between theory and real-world use, and highlights the path forward for robust evaluations of closed and open source models under real-world noise. In "Shallow Robustness, Deep Vulnerabilities: Multi-Turn Evaluation of Medical LLMs," Dynamo's ML Research team: 📉 Found that Anthropic's Claude Sonnet 4 drops from 91.2% to 13.5% accuracy when injected with real-world noise. Similar drops occur for OpenAI and Google models. 🔊 Constructed a new evaluation framework that analyzes the gap between theoretical and real-world performance of LLMs by introducing noisy contextual information to simulate what happens day-to-day.  💻 Released an interactive site, dataset, and code for you to evaluate robustness of closed and open source models under real-world noise. Curious about DynamoEval's multi-turn, context-rich, custom evaluations? Learn more here: https://bit.ly/4pBVLKL 📄 Paper: https://lnkd.in/e4ZHnbXh 📊 Results: https://lnkd.in/eCp86CJn 🗂️ Dataset: https://lnkd.in/eJ942fAS 🧑💻 Code: https://lnkd.in/ej4bwXWc #NeurIPS #NeurIPS2025 #AIEvaluations #AIRobustness #AIResearch

Major Dynamo Milestone: US Army Selects Dynamo AI to Advance Scalable AI Risk Management for Mission-Critical Defense Applications This contract reinforces Dynamo AI’s commitment to ensuring AI systems are robust, secure, and mission-ready for high-impact, cutting-edge deployments. Through this contract, Dynamo will: ✅ Develop tailored AI model evaluations and guardrails for U.S. Army use cases  ✅ Deliver real-time observability, security, and privacy  ✅ Implement automated stress testing to identify risks such as data leakage, adversarial attacks, and hallucinations At project completion, the Army will gain a ready-to-deploy, mission-aligned toolkit that enhances AI risk management, ultimately strengthening the security, reliability, and transparency of AI-enabled systems across critical defense operations. To learn more, see our full announcement here: https://bit.ly/3KSVhk2

Dynamo’s Head of AI Compliance Strategy Daniel Ross recently joined global financial regulators and industry leaders at GFTN’s Insight Forum for panels to discuss AI adoption, governance, and risk management in the financial services industry. In his panels with industry leadership, global regulators, and AI researchers, Dan highlighted three observations that practitioners and government stakeholders should be closely monitoring: 💼 While AI risk controls are increasingly technical, these technologies need to empower LRC and business leaders to stay actively engaged in both control design and monitoring 🤝 Regulator–industry collaboration around best practices, guidance, and operating models for AI governance is essential 🛡️ Risk teams need visibility into actual control strength. Thanks to Global Finance & Technology Network (GFTN) and Monetary Authority of Singapore (MAS) for bringing together such an excellent group for these conversations. We look forward to future contributions. #IF2025 #GFTN #AIRisk #AIGovernance #DynamoAI

Dynamo AI's Chief Product Officer and Cofounder Christian Lau spoke to a crowd of AI and security leaders at Microsoft #Ignite conference in San Francisco. Here's a quick recap in case you missed it! Dynamo announced its integration with Microsoft 365 Copilot, where enterprises can integrate their highly customized Dynamo guardrails into M365 Copilot to detect or block prompts and responses that violate your custom governance policies. Dynamo presented a live demonstration of common guardrails across enterprises, including flagging real-time violations of: 🌐 EU AI Act Prohibited AI activities 🗽 NY State Law 144 on leveraging AI for employment related decision-making ⚖️ Reg B of the Equal Credit Opportunity Act -- Introducing Bias against Protected Classes in Credit Underwriting Tasks with AI Dynamo detailed a comprehensive control framework for securing both AI Agents and 3rd party SaaS AI apps. Thank you to Heena Purohit, Tom Davis and the Microsoft team for the opportunity to share our joint work together enabling secure and compliant adoption of Generative AI!

This Wednesday, Dynamo's Vaikkunth Mugunthan and Daniel Ross will be taking the stage at AI Verify Foundation's final Community Event of the year to share more about our 🧪 custom, automated AI test and evaluation work 🔎 in the Global AI Assurance Sandbox. Will you be in Singapore? Register below to join us! #AIVerify #AITesting #AIGovernance

💡Dynamo AI Speech at Microsoft Ignite 2025⚡ Attending #MicrosoftIgnite? Dynamo AI Co-Founder and Chief Product Officer Dr. Christian Lau will be speaking on: 🔒 Building Trustworthy AI in Financial Services with Azure Christian will explore how financial services institutions can accelerate their generative and agentic AI deployments with the latest advances in evaluations, guardrails, and observability — critical ingredients for secure, enterprise-grade adoption. 🗓 Wednesday, Nov 19, 2025 ⏰ 9:30 – 10:00 AM PST 📍 THR741 Stop by to learn more about Dynamo AI’s recipe for secure, compliant AI. Want to meet up? Lets us know here: https://bit.ly/3LzhC6o #MicrosoftIgnite #AI #FinancialServices #Azure #ResponsibleAI #Governance #AIEvaluation #AIGuardrails #MicrosoftForStartups #MSIgnite

🏦 Insights from Dynamo Event with NCUA Chairman Kyle Hauptman on AI & Governance in Financial Services ⚖️ At #Money2020, Canapi Ventures and Dynamo hosted AI, financial services, and governance leaders to discuss the Future of AI in Financial Services in a special in-person session of the Canapi GenAI Council. Conversations ranged from high-value use cases to the evolving governance and regulatory landscape for AI. We were grateful to have National Credit Union Administration (NCUA) Chairman Kyle Hauptman join to share his thoughts on these topics. Notable highlights from the Chairman’s conversation with Dynamo’s Daniel Ross included: 💡 The opportunities and challenges AI presents across financial services, including credit unions 🤝 The need for FSIs to share real-world AI learnings and innovations with regulators to advance responsible adoption in industry and government 🛡️ Why measurable, testable control outcomes are essential in AI risk management, governance, and oversight As AI transforms the financial system, Dynamo remains committed to helping leading FSIs deploy AI securely and with confidence. A special thanks to Chairman Hauptman, the NCUA team, and the Canapi team for helping make this discussion possible! #AIGovernance #Fintech #AIinFinance #Money2020 #AIinFinancialServices #DynamoAI

🚨 Introducing the first of Dynamo AgentWarden’s AI Security Capabilities: Automated Risk Detection and Evaluation for AI Agents🚨 As AI agents transform how enterprises work, they’re amplifying existing vulnerabilities and creating new ones — security and compliance risks that existing processes, tools, and technologies can’t see or stop. Today, we’re excited to launch the first capability of Dynamo’s AgentWarden: ➡️ Automated Risk Detection and Evaluation for AI Agents. AgentWarden continuously analyzes the MCP tools that agents can access, automatically mapping trajectories, attack paths, and vulnerabilities — giving organizations an x-ray view into their AI systems’ real risk surface. With just a few clicks, security and compliance teams can: ⚡ Detect data exfiltration, prompt injection, and compliance risks 🧭 Visualize agent behavior across MCP environments 📊 Generate actionable, auditable risk reports — in under 5 minutes As AI agents accelerate enterprise productivity, AgentWarden ensures that innovation stays aligned with governance, security, and compliance. And AgentWarden won’t stop here: stay tuned for upcoming releases feature AgentWarden’s custom guardrail and real-time observability capabilities for AI agents. 🔐 Secure the future of agentic AI. Learn more in our overview blog and schedule a demo here: https://bit.ly/4oYc3ga #DynamoAI #AgentWarden #AgenticAI #AgentSecurity #AISecurity #AIGovernance

🏛️ New Blog Recap on Dynamo AI Co-Founder’s Congressional Testimony📜 On September 18, 2025, Dynamo AI Co-Founder and President Dr. Christian Lau testified before the U.S. House of Representatives Financial Services Subcommittee on the Digital Assets, Financial Technology, and Artificial Intelligence. In the testimony, Christian highlighted that the financial services industry is at a critical moment: while AI promises unprecedented opportunities, institutions struggle to navigate unique risks in heavily regulated, high-impact environments. For an overview of the testimony, read our new blog post here: https://bit.ly/4qeoLJ9 His key message was clear: alongside core principles and practices for AI risk management, AI itself can be a critical protective control to mitigate many key AI and security risks in the sector. The path forward requires that the United States lead not just in building powerful AI systems, but in developing the protective infrastructure and policy frameworks that incentivize secure AI deployment at scale. Thanks to HFSC Chairman French Hill, Subcommittee Chairman Bryan Steil, and all Subcommittee members and staff for their leadership on these issues. Dynamo AI is proud to continue to support this important work in AI and risk management policy, in Washington and beyond. Thanks to Forbes for the coverage here: https://bit.ly/3LdArvy

AI is transforming the financial system — and questions around ROI, governance, and the future of regulation are more critical than ever. On the sidelines of #Money2020 at the Venetian’s brand new Bazaar Meat by José Andrés, Dynamo AI and Canapi Ventures will co-host an exclusive side event featuring Kyle Hauptman, Chairman of the National Credit Union Administration (NCUA). Join leading voices across AI, financial services, governance, and policy for coffee, cocktails, and conversation as we explore “The Future of All Things AI in Financial Services” — from high-value use cases to the evolving governance and regulatory landscape. 📅 Tuesday, October 28 | 2–4 PM PT 📍 Outdoor bar and patio at Bazaar Meat by José Andrés, Palazzo at the Venetian Resort  🗣️ Sidelines of Money20/20 👉 Interested in attending? Let us know here: https://bit.ly/474ig41 #AIGovernance #Fintech #ResponsibleAI #AIinFinance #Money2020 #FinancialServices #DynamoAI