New architecture separates reasoning, memory, control, execution and goals to provide a structured record of how AI systems reach their outputs
United States, 8th Oct 2026 — FORHU.Inc is advancing research into transparent artificial intelligence through its Structured Cognitive Loop (SCL), an architecture designed to address a central challenge in modern AI: systems can produce useful answers without providing a sufficiently clear record of how those answers were reached.
As AI adoption expands across business, legal and other high-stakes environments, the ability to understand, verify and audit AI-generated outputs is becoming increasingly important. SCL approaches this challenge by treating transparency as part of the underlying architecture rather than as a reporting feature added after an answer has been generated.
Many conventional AI systems operate as effective but difficult-to-inspect black boxes. Users may receive an answer, recommendation or generated document without being able to independently trace the sequence of reasoning and information that produced it.
SCL, which FORHU describes as a “glassbox” architecture, is designed to record and separate the major steps involved in producing an AI output. This creates a structured trail that can be reviewed, verified and audited, helping organizations better understand where an answer came from and identify potential problems.
SCL organizes AI activity into five interconnected modules modeled on broad aspects of human cognition. Judgment handles reasoning and evaluation; Memory manages accumulated experience and relevant information; Control supports self-monitoring and oversight; Runtime manages execution; and Metaprompt defines goals, constraints and operating requirements.
The separation is intended to make AI processes easier to inspect. Rather than treating a final response as the only meaningful output, SCL records the intermediate structure that leads to it, creating a clearer basis for review and accountability.
FORHU also applies SCL to the problem of AI hallucinations, in which an AI system generates information that is inaccurate, unsupported or fabricated. The company does not position hallucinations as a problem that can simply be eliminated through a single technical mechanism.
Instead, SCL is designed to record, verify and correct mistakes as part of an ongoing governance process. Its guiding principle is summarized by the motto: “the right to err, the duty not to repeat.” The objective is to make errors identifiable and actionable rather than allowing the same unsupported output to pass through the system without a traceable record.
FORHU’s 2026 research introduces Pool-Gated Retrieval, a method designed to address situations in which an AI system lacks sufficient information to answer a question reliably. The approach is intended to prevent the system from filling information gaps by inventing an answer.
Instead, retrieval is constrained by an available evidence pool, while the system maintains an auditable record of the evidence associated with its response. This provides a mechanism for distinguishing between information supported by available evidence and situations where the system should recognize that sufficient information is unavailable.
FORHU’s research on transparent AI has been published through platforms including arXiv, PhilSci Archive, OSF and Preprints.org. The company also showcased its GlassboxAI work at VivaTech 2026 in Paris, providing an opportunity to present its approach to transparency and AI accountability to an international technology audience.
The architecture is particularly relevant as regulators and businesses examine requirements surrounding explainability, documentation, risk management and accountability. The European Union’s AI Act has increased attention on how AI systems are developed, documented and governed, particularly when they are used in higher-risk contexts. SCL is intended to contribute to this broader effort by making traceability part of the system architecture.
“Trust in AI cannot be created simply by adding explanations after an answer has already been produced,” said Jungkwan Shin, CEO of FORHU.Inc. “Transparency must be built into the structure of the system itself, so that reasoning, evidence, constraints and decisions can be examined rather than merely accepted. Our goal with SCL is to provide a practical foundation for AI that can be understood, reviewed and held accountable.”
FORHU is applying SCL research to practical use cases, including ilovelawyer, a legal AI web application where verifiable reasoning and evidence are particularly important. The project is intended to demonstrate how structured AI processes can support users in contexts where understanding the basis of an AI-generated response is essential.
FORHU stands for “FOR HUMAN.” The company believes AI should expand human potential, not replace human judgment. FORHU.Inc is headquartered at 30 Wall Street, 8th Floor, New York, NY 10005.
Contact FORHU.Inc:
Jungkwan Shin, CEO
Email: support@forhu.ai
Website: https://www.forhu.ai
Research: https://www.forhu.ai/research
Organization: FORHU.Inc
Contact Person: Jungkwan Shin, CEO
Website: https://www.forhu.ai/
Email: Send Email
Country: United States
Release id: 49783
The post FORHU Advances Glassbox AI Architecture With Structured Cognitive Loop to Make AI Reasoning Traceable and Auditable appeared first on King Newswire. This content is provided by a third-party source.. King Newswire is a press release distribution agency. We do not accept any responsibility or liability for the accuracy, content, images, videos, licences, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright concerns related to this article, please contact the company listed in the ‘Media Contact’ section above.
Disclaimer: The views, recommendations, and opinions expressed in this content belong solely to the third-party experts. This site was not involved in the writing and production of this article.