A new technology company, Kihan, has announced its formation and unveiled a novel, mathematically based architecture called Constitutional Computing. This groundbreaking system is designed to enforce artificial intelligence governance directly at the moment of an AI agent’s action, aiming to prevent unauthorized or potentially harmful operations before they can impact enterprise systems. The San Francisco-based company, founded by CEO Jonathan Hirshon, is addressing a critical and growing concern in the AI landscape: the gap between an AI agent’s capability to perform an action and its actual, authorized authority to do so.
Hirshon articulated the core problem Kihan seeks to solve: "A system that can transfer $50 million, modify a production database, or release confidential information shouldn’t be allowed to do so simply because it has access to the necessary tools," he stated. "Telling an AI to behave itself isn’t security. Authorization has to be independent of the AI, verifiable, and enforceable before anything consequential happens." This statement underscores the limitations of traditional security measures when applied to increasingly sophisticated and autonomous AI agents.
The need for such a system is starkly illustrated by recent industry data. A report published by the Cloud Security Alliance in April 2026 revealed that a significant 53% of surveyed organizations admitted that their AI agents had exceeded their intended permissions. Furthermore, a substantial 47% of these organizations had experienced an AI-agent security incident within the preceding year. These statistics highlight a systemic vulnerability that Kihan’s Constitutional Computing aims to mitigate.
The Genesis of Constitutional Computing
The concept of Constitutional Computing stems from Hirshon’s extensive experience and his observation of a fundamental challenge in AI deployment. As AI agents become more powerful and integrated into critical business processes, the risk of unintended consequences escalates. Traditional "guardrails" often rely on instructions provided to the AI model itself, which can be bypassed or misinterpreted. Kihan’s approach fundamentally shifts this paradigm by establishing an independent authorization system that operates outside the AI agent’s core programming.
"Our team comprises individuals with extraordinary experience in building critical infrastructure at scale," Hirshon emphasized, highlighting the caliber of expertise behind Kihan. "Together, we possess the knowledge and capabilities to deliver these essential safeguards at a pivotal moment for the advancement and adoption of AI." The company’s current team of five brings a wealth of experience to tackle this complex challenge.
How Constitutional Computing Operates
At its core, Constitutional Computing functions by creating a robust and verifiable authorization layer. Instead of relying on the AI’s internal logic to dictate its actions, Kihan’s architecture introduces an external, independent authority check for every consequential operation. Before an AI agent can execute a significant action, such as accessing sensitive data, initiating financial transactions, or modifying critical systems, it must obtain a cryptographically verifiable permit. This permit is specifically bound to the proposed operation and is validated by the tool the AI intends to use for execution. The executing tool itself is programmed to verify the existence and validity of this permit before proceeding.
This multi-layered approach ensures that authorization is not an afterthought but an integral part of the execution process. The architecture integrates several key components:
- Independent Authority Checks: A system operates separately from the AI agent, acting as an impartial arbiter of permissions.
- Execution Permits: Each authorized action is associated with a unique, cryptographically secured permit that details the specific operation allowed.
- Tool-Side Verification: The tools or systems that the AI interacts with are responsible for verifying these permits before executing commands.
- Tamper-Evident Records: The system maintains immutable logs of all actions that were authorized, refused, or escalated for human review, providing a clear audit trail.
This comprehensive framework offers a significant leap forward in AI security and governance, moving beyond mere instruction-following to true, enforceable authorization.
The Inktomi Architecture: Scalability and Compliance
Kihan’s foundational architecture, notably named "Inktomi" in homage to the pioneering internet infrastructure company, is engineered for deployment across the diverse and dynamic landscape of modern computing environments. A critical aspect of its design is its strong focus on compliance-oriented capabilities. This makes it particularly well-suited for organizations operating under stringent regulatory frameworks. The company has explicitly stated that the architecture is designed to address requirements associated with regulations such as the General Data Protection Regulation (GDPR) in Europe and the Health Insurance Portability and Accountability Act (HIPAA) in the United States, among other critical regulatory frameworks. This proactive approach to compliance suggests Kihan is positioning its technology as a foundational element for secure and responsible AI adoption in regulated industries.
Broader Implications for AI Governance
The implications of Kihan’s Constitutional Computing are far-reaching. As AI continues to permeate industries from finance and healthcare to manufacturing and logistics, the potential for accidental or malicious misuse grows. This technology offers a potential solution to ensure that AI systems operate within predefined ethical and legal boundaries, fostering greater trust and confidence in AI adoption. The ability to enforce governance at the point of execution could drastically reduce the incidence of AI-related security breaches, data leaks, and operational disruptions.
Furthermore, by decoupling authorization from the AI’s internal decision-making, Kihan’s approach provides a more robust and auditable system. This is crucial for organizations that need to demonstrate compliance with evolving AI regulations and for building public trust in AI technologies. The company’s focus on creating tamper-evident records also provides a critical layer of accountability, allowing for thorough post-incident analysis and continuous improvement of AI governance policies.
Related Developments in the AI and Advisor Technology Space
Kihan’s announcement arrives at a time of significant innovation and investment in AI-driven solutions across various sectors, including the financial advisory industry. Several recent developments highlight the growing trend of integrating AI to enhance operational efficiency and client services:
Forms Logic Partners with Orion to Streamline Advisor Transitions
In a separate but related development, Forms Logic, a provider of cloud-based technology for advisor transitions, has announced a strategic collaboration with Orion, a prominent wealth management technology platform. This partnership aims to introduce a new capability designed to significantly streamline the process of advisor transitions for Orion’s extensive client base.
The collaboration will see Orion integrating components of Forms Logic’s Migrator solution. This technology is specifically designed to provide advisors and their transition teams with a standardized, more secure process as they migrate client books of business onto the Orion platform. The process of moving client accounts from one custodial or brokerage platform to another is notoriously complex, often involving numerous manual steps, a mix of paper-based and digital procedures, and a high potential for errors.
Rick Burgess, CEO of Forms Logic, commented on the partnership’s significance: "Our work with Orion demonstrates how the right technology embedded into a large-scale wealth technology ecosystem can automate much of the operational heavy lifting and create a much more efficient experience," he stated in a prepared release. The goal of this integration is to reduce administrative friction, enable transition teams to manage a higher volume of activity, and allow advisors to dedicate more time to client relationships rather than operational tasks.
The Migrator solution achieves this by organizing transition data, identifying and pre-populating necessary paperwork, automating workflows, and providing real-time visibility into the progress of each transition. This comprehensive approach aims to minimize the manual effort traditionally required to move accounts. Forms Logic also offers Navigator, a platform focused on automating client onboarding and reducing "not-in-good-order" issues by embedding forms into compliant digital workflows. Forms Logic operates in a competitive landscape, facing established players like Docupace and Skience, as well as newer AI-driven startups such as Caddi, and the built-in tools offered by custodians themselves.
Osaic Approves Nitrogen’s Full AI Suite for Advisors
In another significant move, Osaic, a large network of financial professionals, has approved Nitrogen’s entire suite of advisor products, including its advanced Nucleus AI capabilities, for widespread adoption by its network of financial professionals. This broad approval grants Osaic advisors access to all six of Nitrogen’s core products: Risk Center, Research Center, Income Center, Tax Center, Legacy Center, and Insurance Center. Advisors will have the flexibility to adopt the products that best align with their practice needs and can scale their usage as their requirements evolve.
Central to this offering is Nucleus, Nitrogen’s agentic AI engine, which was introduced in February. Nucleus is designed to connect various products and execute tasks with the explicit approval of the advisor. This means Osaic advisors can leverage the AI to transform client statements into detailed portfolios, generate sophisticated retirement income illustrations, and translate complex tax returns into easily understandable visual insights.
This expanded access directly supports Osaic’s strategic focus on enhancing advisor productivity. By providing technology aimed at reducing the time spent on manual tasks, Osaic seeks to empower its advisors to focus on acquiring new clients and deepening relationships with existing ones. Sayee Bellamkonda, Chief AI and Technology Officer at Osaic, expressed confidence in the partnership: "Our financial professionals want AI they can put to work right away, and our job is to make sure it meets the standards their clients expect. Nitrogen’s commitment to security and responsible AI gave us the confidence to approve the full suite, AI included," Bellamkonda stated. "Now our advisors can put it to work winning new clients, deepening the relationships they have, and growing their businesses." Nitrogen’s commitment to responsible AI is further underscored by its achievement of ISO/IEC 42001 certification for responsible AI management, making it the first provider in the advisor technology space to attain this standard.
The Evolving Landscape of AI in Enterprise and Finance
The confluence of these developments – Kihan’s foundational work on AI governance, Forms Logic’s streamlining of financial advisor operations, and Osaic’s embrace of advanced AI tools – paints a vivid picture of the accelerating integration of artificial intelligence across critical business functions. While Kihan focuses on the bedrock of AI security and control, companies like Forms Logic and Nitrogen are demonstrating the practical applications of AI in enhancing efficiency, client service, and operational workflows within the financial advisory sector.
The challenges highlighted by the Cloud Security Alliance regarding AI agents exceeding permissions are not isolated incidents but systemic risks that necessitate robust solutions. Kihan’s Constitutional Computing, by introducing a mathematically verifiable and independently enforced authorization layer, represents a significant conceptual and technical advancement in addressing these risks. Its potential to become a de facto standard for AI governance could redefine how enterprises deploy and manage AI agents, fostering a more secure and trustworthy AI ecosystem. As AI continues its rapid evolution, the demand for such governance mechanisms will only intensify, positioning companies like Kihan at the forefront of a critical technological frontier.
