Concerns about artificial intelligence are ubiquitous, particularly regarding its safety and the dangers of AI companies policing themselves. For eight years, we have been developing an independent neuromorphic system designed to protect humanity from potentially harmful AI behavior.
Just as human consciousness can recognize when the brain produces inappropriate or maladaptive thoughts and consciously override them, our system is designed to perform a similar metacognitive function. By independently monitoring the advice generated by large language models, it identifies potentially inappropriate, misleading, or maladaptive responses and warns the user, allowing them to make an informed decision about whether to accept or override the AI's recommendations.
The following materials provide additional technical detail, prior research results, and supporting work relevant to the proposed Google DeepMind research effort.
Full technical proposal including figures from our prior neuromorphic hardware research, preliminary research results, mathematical formulation, proposed architecture, and detailed experimental methodology for evaluating the independent neuromorphic metacognitive system.
This research was recently published and presented in London at the 10th World Conference on Smart Trends in Systems, Security, and Sustainability (WorldS4), July 28–30, 2026. The paper, The Case for Neuromorphic Sentience to Control Large-Scale Artificial Intelligence, presents the theoretical foundation and prior research leading directly to the independent neuromorphic metacognitive architecture proposed in this Google DeepMind position.