NewMind AI ads

Lehman Brothers: Don’t Tell Me Why It Collapsed. Show Me the Evidence. When investigating a failure as complex as Lehman Brothers, the challenge isn’t simply finding documents. It’s connecting thousands of pieces of evidence into a defensible chain of events. In this demo, we put nmaistro to the test. Did Lehman make its balance sheet appear stronger through Repo 105 transactions? When did management understand the severity of the liquidity crisis? Did the risk profile presented to investors match what was being reported internally? Which warnings appeared in the final 90 days before the collapse—and what happened after they were raised? Instead of asking nmaistro to speculate, we ask it to reconstruct the story from the underlying evidence: accounting records, management presentations, emails, liquidity reports, board materials, SEC disclosures, investor communications and internal risk reports. The result is not just an answer. It is a reconstruction of the events behind the collapse—connecting claims to documents, documents to events, and events to a timeline that can be examined and verified. From millions of pages to the evidence that matters. #nmaistro #NewMindAI #AgenticAI #LehmanBrothers #LegalAI #Evidence #FinancialAnalysis

From Investing in a Bank to Controlling One Acquiring shares in a bank may look like a corporate transaction. But once ownership begins to translate into influence and control, the regulatory picture changes significantly. In our latest nmaistro use case, we put Bank Compliance MCP to work on two connected questions: When does the acquisition of shares in a Turkish bank require regulatory permission? And what new obligations and risks arise when that investor becomes the bank’s controlling shareholder? What begins as a question about share acquisition quickly develops into a much broader examination of banking regulation, ownership, control and regulatory responsibility. With Bank Compliance MCP, nmaistro enables legal and banking professionals to investigate these interconnected questions across specialized banking resources and trace the analysis back to the relevant legal sources. Because in banking, the real question isn’t always what you can acquire. It’s what comes with it. #nmaistro #BankCompliance #BankingLaw #BankingRegulation #Compliance #LegalAI #FinancialRegulation

We are pleased to share that NewMind AI has been awarded 100,000 GPU hours at the Barcelona Supercomputing Center (BSC) through EuroHPC Joint Undertaking (EuroHPC JU). This allocation will support the next phase of our research into reliable, traceable, and professionally deployable AI systems for the legal domain with a particular focus on enabling AI agents to perform complex legal tasks with greater consistency, transparency, and accountability. A core part of this research is the development of ontology driven AI systems that can better understand legal concepts, relationships, and professional context while producing outputs that remain interpretable, source-traceable, and auditable. Building on our previous large-scale EuroHPC research, we now have the opportunity to continue advancing this work at significant computational scale. We look forward to putting these 100,000 GPU hours to work. We extend our sincere thanks to EuroHPC for recognizing the value of our research and entrusting us with the computational resources to take it further. #NewMindAI #EuroHPC #BarcelonaSupercomputingCenter #BSC #LegalAI #AIResearch #Ontology #Supercomputing

Operational Evidence #1 — Argument & Evidence Mapping One of the recurring questions we encounter is whether AI can move beyond summarization and become genuinely useful in complex legal work. Our answer is that usefulness begins when a system can identify, structure, and operationalize legal knowledge. In this demonstration, based on the Metalclad arbitration, the objective is not to generate a narrative summary of the case. Instead, the system identifies the arguments advanced by the parties, extracts the evidence associated with each argument, and maps the relationships between claims, facts, documents, procedural events, and legal positions. What makes this possible is not merely retrieval, prompting, or larger language models. The underlying architecture is ontology-driven and reflects what we describe as Agentology: the design of operational environments in which AI agents work within defined semantic jurisdictions, governed by structured knowledge, relationships, rules, and responsibilities. Thousands of documents are transformed into a coherent semantic environment where arguments, evidence, legal concepts, actors, procedural events, and documentary records become part of an operational knowledge structure. The result is not information consumption. The result is operationalization. This is what we mean by Operational Evidence: not theoretical claims about AI capabilities, but observable demonstrations of how ontology-driven and agentological systems can transform large-scale legal records into actionable legal intelligence. Because the future of legal AI is not about producing more text. It is about making complex legal knowledge executable. #Agentology #Ontology #LegalAI #AgenticAI #KnowledgeArchitecture #LegalTech #Arbitration #ExplainableAI #KnowledgeGraphs #OperationalEvidence

Operational Evidence #2 — Reconstructing the Story of a Case At London LegalTech Talk, one of the use cases we demonstrated focused on a challenge familiar to every arbitration and litigation lawyer: reconstructing the chronology of a dispute. This is the second use case from that workshop, and the video below serves as an Operational Evidence of that capability. In complex disputes, understanding a case requires far more than retrieving documents or generating summaries. A meaningful chronology emerges only when events, actors, documents, contractual relationships, procedural developments, and legal positions are correctly identified and connected within their broader context. In other words, understanding requires semantics before it requires answers. This is precisely where ontology-driven systems become important. At London LegalTech Talk, we emphasized a principle that sits at the core of our Agentology approach: the quality of an AI agent is ultimately determined by the environment in which it operates. Without a task-specific ontology and a well-defined semantic environment, an agent is often left to infer relationships that have never been explicitly modeled. In complex legal matters, this can easily turn reasoning into sophisticated guessing. Our approach is built upon a semantic architecture that enables the system to identify not only documents, but also the relationships that exist between parties, events, claims, evidence, decisions, and procedural milestones across the entire corpus. Without this semantic structure, AI risks becoming little more than a sophisticated summarization engine. With it, AI can begin to reconstruct the actual story of a dispute. In the Metalclad arbitration, the system reconstructs both the factual evolution of the dispute and the procedural evolution of the case, transforming thousands of pages of materials into a coherent operational representation of legal reality. This is why we refer to this demonstration as Operational Evidence. Not because the system can talk about chronology, but because it can actually produce it. #LegalAI #Agentology #Ontology #SemanticAI #Arbitration #InternationalArbitration #KnowledgeArchitecture #LegalTech #OperationalEvidence #LondonLegalTechTalk

Operational Evidence #3 — Could We Have Seen Lehman Coming? Lehman Brothers' collapse is a defining event of modern financial history. Yet, one question lingers: Was the issue a lack of information, or understanding? Before the collapse, vast public information existed: regulatory filings, analyst reports, news, and more. The challenge wasn't collecting information, but connecting it. At London LegalTech Talk, we showed a use case built on a corpus of public Lehman Brothers data. The goal wasn't claiming the crisis was perfectly predictable, but exploring: If scaled semantic and ontology-driven systems existed, could emerging risks and hidden patterns have been visible earlier? This reflects our Agentology approach: Major events aren't caused by a single fact, but emerge from networks of interconnected actors, decisions, and signals. Without a semantic environment, information is fragmented. Without a task-specific ontology, AI infers unmodeled relationships. But with the right semantic architecture, information becomes an operational understanding of reality. This isn't about hindsight. It's about why AI's future depends less on generating answers and more on understanding relationships. TR Operasyonel Kanıt #3 — Lehman'ın Gelişini Görebilir miydik? Lehman Brothers'ın çöküşü modern finans tarihinin belirleyici olaylarındandır. Peki sorun bilgi eksikliği miydi, yoksa anlama eksikliği mi? Çöküşten önce muazzam miktarda kamuya açık bilgi mevcuttu: yasal bildirimler, raporlar, haberler ve fazlası. Asıl zorluk bilgiyi toplamak değil, birbirine bağlamaktı. London LegalTech Talk'ta, tamamen kamuya açık Lehman verilerinden oluşan bir senaryo sunduk. Amacımız krizin kesin tahmin edilebileceğini iddia etmek değil, şunu incelemekti: Ontoloji odaklı ve semantik sistemler geniş ölçekte olsaydı, gizli riskler ve örüntüler daha erken fark edilebilir miydi? Bu, Agentology yaklaşımımızı yansıtıyor: Büyük olaylara tek bir gerçek neden olmaz; aktörler, kararlar ve sinyaller ağından doğarlar. Semantik bir ortam olmadan bilgi parçalı kalır. Göreve özgü bir ontoloji olmadan yapay zeka, modellenmemiş ilişkileri türetmek zorunda kalır. Doğru semantik mimariyle bilgi, gerçekliğin operasyonel bir kavrayışına dönüşür. Bu video, geriye dönük bakışla ilgili değil. Yapay zekanın geleceğinin neden sadece cevap üretmeye değil, ilişkileri anlamaya bağlı olduğuyla ilgilidir. #OperationalEvidence #LehmanBrothers #Agentology #Ontology #SemanticAI #KnowledgeGraphs #ArtificialIntelligence #LondonLegalTechTalk