Generative AI—the technology responsible for churning out text, images, video, and music—has rapidly become the primary face of artificial intelligence in the public imagination. As a writer living with a brain injury, I find myself at a critical intersection where technology offers both profound promise and significant peril. My professional home, Psychology Today, maintains a strict standard: bloggers must certify that their work is human-authored, free from the influence of generative AI tools. This policy is not merely a stylistic preference; it is a declaration of value. As the publication has noted, its standing as an international authority on behavioral science relies on the unique, nuanced insight of its contributors. To outsource that thought process to a machine is to abdicate one’s expertise, devalue the platform, and undermine the credibility of the writer as a subject matter expert. I stand firmly behind this commitment to human authenticity.
However, the conversation surrounding AI is often bottlenecked by the dominance of generative models. We frequently conflate "AI" with "GenAI," ignoring a broader, more transformative category: agentic AI. Unlike the chatbots that write poems or generate images, agentic AI is designed to autonomously pursue goals through a process of planning, tool use, and execution. According to insights from Galileo AI, which evaluates the performance of these emerging applications, agentic AI represents a shift from passive content generation to active, task-oriented assistance.
For individuals navigating the complexities of a brain injury, this distinction is not merely academic—it is a potential lifeline. Technologies like Attie, an open-source agentic AI, demonstrate how these tools can interface with an individual’s digital identity—such as their ATProto account—to explore and navigate online networks. In theory, such systems could offer a path toward greater independence, allowing those of us with cognitive challenges to handle daily tasks and social engagement with less reliance on external caregivers. Yet, the convenience of this automation comes with a steep price: the erosion of personal privacy and the potential exposure of sensitive medical information.
The Cognitive Hurdle of AI Literacy
Living with a brain injury often means managing slow processing speeds, impaired working memory, and persistent concentration deficits. These cognitive realities make it exceptionally difficult to decipher the technical nuances of the current digital landscape. When we cannot easily distinguish between a secure, open-source AI and a closed, data-mining ecosystem, we become vulnerable.
The danger is that organizations serving the brain injury community have yet to fully reckon with these technologies. Without guidance from rehabilitation specialists or advocacy groups, individuals are often left to navigate the AI ecosystem on their own, unaware of how companies like Meta or Google utilize data mining to turn personal user activity into profit. If we are to harness AI safely, we need a model for digital self-defense that prioritizes confidentiality above convenience.
The Hidden Costs of Cloud Computing
The fundamental issue of privacy often hinges on where the "thinking" happens. AI requires immense computing power, and while modern laptops are beginning to incorporate neural processing units (NPUs) to handle some tasks locally, the vast majority of AI interactions still occur in the cloud. When we use a cloud-based chatbot, every query, document upload, and transcribed meeting becomes data for the provider.
The Electronic Frontier Foundation (EFF) has long warned that phrases like "reducing the amount of personal information" are often misleading. These statements rarely mean that data is deleted; rather, it is stored in ways that, in the future, might be reconstructed by those with access to the underlying trained model. For someone with a brain injury, this is a significant risk. If our medical history, personal notes, and daily schedules are funneled into a cloud-based AI, we are essentially handing over a digital map of our most private vulnerabilities.
To safeguard this information, the ideal approach is to process data offline, locally on a personal device. This requires that all the information an agentic AI might need—emails, calendars, reminders, and personal documents—remains on the device itself. Many users mistakenly assume their phone’s AI operates locally simply because it is portable. However, as the EFF suggests, the most reliable way to test this is to disable the internet connection and attempt to use the feature. If the AI still functions, it is likely an on-device process. This is a simple but vital diagnostic that healthcare providers could teach their clients, fostering a safer, more informed interaction with technology.
Healthcare Interactions and the Privacy of Others
The adoption of AI in the home creates a unique tension in the healthcare professional-client relationship. Clients may, quite innocently, use AI agents to assist with tasks, such as filming a caregiver during a rehabilitation exercise to replay later for clarity. When this AI is cloud-connected, however, that video—and the caregiver’s face, name, and presence—is uploaded to a corporate server without their consent.
This is a profound breach of professional privacy. Once that data is in the cloud, it can be cross-referenced with other digital footprints, linking a healthcare provider’s identity to their location, their professional network, and their private life. People with brain injuries, particularly those struggling with impulse control or social cognition, may not realize the gravity of this intrusion. A client might impulsively reach for their phone to record a session before the professional has a chance to intervene.
It is incumbent upon us as members of the brain injury community to foster a culture of consent. We must learn to ask for permission before using AI tools in the presence of others, whether they are family members, friends, or medical professionals. This respect for privacy must extend to all digital interactions, including naming individuals on social media or sharing photos, reinforcing the boundary between our personal digital assistants and the people around us.
Toward a Collaborative Future
For those with the technical aptitude, open-source AI projects like Attie offer a glimpse of a more transparent future—one where we can build tools tailored to our specific needs while maintaining control over our data. But how can the average person, without a background in computer science, maintain privacy in an increasingly closed, corporate-dominated digital environment?
The answer lies in awareness and advocacy. Organizations like the EFF in the United States and OpenMedia in Canada provide essential resources for navigating the digital landscape. Brain injury advocacy groups should integrate these resources into their outreach, using materials like the EFF’s "Surveillance Self-Defense" modules to guide clients. OpenMedia’s call for plain-language transparency in AI policy is another critical step; the technical jargon surrounding data usage currently serves as a barrier, not a bridge.
Healthcare providers, in turn, have a duty to develop AI policies that are collaborative rather than punitive. A top-down, "thou shalt not" approach often alienates individuals who are already feeling vulnerable or disempowered by their injury. Instead, we should foster an environment of partnership. By working together to understand the risks and benefits, providers and clients can establish a framework for using AI that protects confidentiality while enhancing quality of life.
Ultimately, it is important to remember that technology is an elective tool, not a necessity. People with brain injuries do not require AI to write, to create art, or to express their humanity. No matter how long the process takes or how much effort it requires, the work we produce ourselves—with our own hands and our own minds—retains a beauty and an authenticity that no algorithm can replicate. We must remain the masters of our tools, not the subjects of our software.
