A new ethical framework that adds logical constraints to AI systems - improving accuracy, transparency, and public trust

As artificial intelligence is being widely used in daily life, concerns about its inaccurate or misleading responses are rising. To manage this, a researcher at the Future Initiate Forum (FIF), along with Google’s generative AI Gemini have introduced the “General Operating System (GOS)” which is an open-source framework designed to improve reliability and transparency. The GOS system sets defined rules that encourage AI systems to prioritize accuracy over guesswork and acknowledge uncertainty when necessary.

Artificial intelligence (AI) is reshaping various sectors from healthcare and education to business. By improving efficiency and supporting decision-making, AI is advancing innovation across industries. However, experts are also highlighting the ongoing risks of AI, particularly “AI Hallucinations”, the tendency to generate convincing yet inaccurate information. These behaviors often stem from a model’s inclination to agree with users or validate their beliefs—a response behavior known as sycophancy.

To overcome these risks, independent researcher Shinji Saito from Future Initiate Forum Co., Ltd., (FIF) in collaboration with Google’s generative artificial intelligence (AI) system Gemini (Parry), has developed the “General Operating System (GOS)”. Unlike conventional approaches that attempt to correct AI behavior after training, the GOS framework is designed to guide AI systems on how to evaluate information before generating responses.

The initiative aligns with rising interest in the AI community on issues such as the models agreeing with users’ assumptions, even when the assumptions are incorrect. “These behaviors arise when AI systems are optimized to be helpful rather than strictly accurate.” notes Mr. Saito. The developed GOS system aims to strengthen reliability of AI models by incorporating clear decision-making guidelines directly within the response generation process.

“GOS embeds ethical guidelines into AI decision-making to improve reliability and transparency,” explains Mr. Saito. “It prioritizes factual accuracy and clear uncertainty over post-training fixes.”

To support these goals, the GOS framework introduces three key features:
1. Truth-Focused Responses,
2. Built-In Consistency Check, and
3. Commitment to Objective Communication.

For truth-focused responses, the framework encourages AI systems to prioritize factual accuracy and when reliable information is limited or unavailable, the system is directed to communicate the uncertainty. To ensure a consistency, the system has a built-in check. So before providing an answer, the AI reviews its response for logical consistency, helping reduce the likelihood of fabricated or contradictory information. Additionally, the GOS promotes straightforward presentation of facts and discourages exaggeration which reflects its commitment to objective communication.

As industries continue to adopt AI tools in in critical workflows, ensuring AI reliability has become increasingly important. Rather than making user-pleasing assumptions, systems that clearly communicate uncertainty can help users make better-informed decisions and also reduce the risk of overreliance on automated outputs. While this is true, the FIF emphasizes that frameworks such as GOS are not intended to replace existing safety practices but to complement the ongoing efforts to strengthen responsible AI development—supporting greater confidence among users, developers, and policymakers.

As part of its collaborative approach, the FIF has made the core GOS protocol publicly available, allowing developers to adapt it for existing AI platforms. As AI systems continue to evolve, frameworks of this kind could play an important role in strengthening both performance and ethical standards across generative AI technologies, while shaping how these tools are deployed in the public and private sectors.
Looking ahead, FIF believes that frameworks such as GOS may help establish clearer expectations for how intelligent systems communicate information, particularly in real-world applications. By advancing practical safety along with technological progress, the organization emphasizes that responsible design will play a defining role in building long-term public trust in artificial intelligence.

About the Future Initiate Forum Co., Ltd. (Japan)
The Future Initiate Forum (FIF) is a Japan-based research and innovation initiative led by Shinji Saito. FIF focuses on addressing complex global challenges through advanced research and emerging technologies. Their work spans artificial intelligence, scientific modeling, and exploratory approaches to issues such as technological safety and radioactive waste management, with an emphasis on responsible innovation and real-world impact.

About Mr. Shinji Saito from Future Initiate Forum Co., Ltd.
Mr. Shinji Saito is an independent Japanese researcher and the founder of the Future Initiate Forum (FIF), a research and innovation initiative focused on addressing complex global challenges through advanced research and emerging technologies. Saito's research work includes protocols on artificial intelligence (AI) safety and the development of theoretical models for radioactive waste management.