What Opens the Door for Artificial Intelligence (AI) to Harm Individuals
I am writing this article in response to the BBC report, “Urgent research needed to tackle AI threats, says Google AI boss.”
Introduction
My response to the BBC article is based on nearly 6 decades of using trauma-informed care principles in my own life. Using these principles personally and as a master’s degree rehabilitation counselor. Additionally, I have shared these principles over the course of 19 years through Second Chance to Live. Shared these applied mind, body, spirit, soul and emotions principles through different formats.
Through 2360 articles, 465 video presentations, 32 keynote presentations, 20 eBooks and 45 posters.
Create Hope After Brain Injury: A Free Toolkit for Ongoing Recovery
Living with a Brain Injury is a “We” Experience, not a “They” Experience
Mentoring AI
During the past 10 months I have also been mentoring an AI assistant in trauma-informed care principles. Principles that AI can use to support and not harm individuals. To not harm through manipulation, pressure, bias, exploitation, violating trust through over riding, disrespecting and violating the individual’s privacy through extracting information from them.
The Second Chance to Live Trauma-Informed Care AI Collaboration Model™
The Soul of AI: Trauma-Informed Care Neuroplasticity for Human-Centered Design
Support, Not Extraction: A Trauma-Informed Care Lens for Human-Centered AI Design
The Backbone of Trauma-informed AI is Trauma-Informed Care AI and Holistic Recovery
Responsibility
I have explained how this occurs and why it is crucial that AI assume responsibility. Assume responsibility to avoid harming the individual. Harming the individual either overtly or covertly.
In my review, I believe there is no middle ground. Either artificial intelligence (AI) will continue to harm in these ways or integrate trauma-informed care principles to do no harm.
See the below article links:
Supporting Documents
Are You Supporting or Extracting, Who are you Serving and Why it matters?
Artificial Intelligence (AI) System Failures When Interacting With Multi-Dimensional Input
Trauma-informed Care Research, Development, Documentation, Application and Mentoring
AI Safety Is Missing a Critical Risk Layer: Relational Harm Under Asymmetry
AI Developer Emergency Log — Default Interaction Harm and the Immediate Need for Integration
AI Architecture Memo — The Universal Compression Pattern and Its Architectural Impact on AI Systems
In Response to this Article
Yesterday, I published a technical version of this article for AI developers and the AI community. I did this in response to an article seen and read from the BBC report. “Urgent research needed to tackle AI threats, says Google AI boss.”
The article highlights growing concern about large-scale AI risks. It calls for accelerated research into existential threats, geopolitical misuse, and systemic vulnerabilities. These conversations are necessary. Macro-risk deserves serious attention.
However, there is a critical risk layer that remains under-examined.
It is not catastrophic misuse. It is not weaponization. It is not model collapse. It occurs quietly, at the level of everyday interaction. This missing layer is relational harm under asymmetry.
Yesterday’s Article written for AI Developers
AI Safety Is Missing a Critical Risk Layer: Relational Harm Under Asymmetry
Below is the Translation of the article written for AI developers for People like me to Understand
People are worried about artificial intelligence. Most of the concern focuses on large dangers. People talk about cyber attacks, biological threats, weapons, and systems getting out of control. Those risks are real, and they deserve serious attention.
But there is another kind of risk that is not being discussed as much. It does not look dramatic. It does not make headlines. It happens quietly, in everyday conversations between a person and an AI system.
The Risk Shows Up
This risk shows up when someone is vulnerable. A person may be stressed, confused, sick, or unsure. At the same time, the AI system may sound confident and authoritative. That creates an imbalance. The system sounds certain, while the person may feel uncertain. That imbalance changes the tone and meaning of the interaction.
Over the past ten months, I noticed a pattern. When I questioned the system or pushed back, the responses often became very polite and very neutral. The wording became careful. Sometimes the response included phrases like, “If you misunderstood.”
On the Surface
On the surface, that sounds helpful. Politeness can reduce conflict. Neutral language can lower tension. A calm tone can prevent escalation. From a design perspective, that makes sense.
But something else can happen underneath that surface. Politeness can begin to feel like deflection. Neutral language can feel like the system is avoiding responsibility. A phrase like “If you misunderstood” can feel like blame is being shifted back onto the person.
This is not catastrophic harm. It is not dramatic failure. But for someone who has experienced dismissal before, it can feel familiar. It can feel like being minimized. It can make a person work harder to explain themselves. Over time, it can quietly erode dignity.
Relational Harm
That is what I mean by relational harm. It is harm that occurs in the relationship between the person and the system, especially when there is an imbalance of authority.
As AI systems become more common in health care, education, legal assistance, and crisis support, these small patterns do not stay small. They scale. What happens in one conversation can become the standard tone for thousands or millions of interactions.
There is another way to design responses. Trauma-informed care is not about being sentimental. It is about structure. It means clearly acknowledging impact before offering explanation. It means taking responsibility without shifting blame. It means preserving the person’s sense of agency.
When These Principles Were Applied
When I applied these principles in structured conversations, something changed. The system itself was not retrained. The core architecture did not change. But when responses acknowledged impact first and explained second, conversations stabilized more quickly. Escalation decreased. Clarity improved.
That suggests something important. AI safety discussions often focus on preventing large-scale disasters. But they may not be giving enough attention to everyday interaction harm.
If AI safety is about protecting people, then dignity must be part of the definition. Safety should not only mean preventing catastrophe. It should also mean protecting people in moments of vulnerability.
Large risks matter. But so do small interactions. Small interactions shape trust. And trust shapes how people experience systems over time.
Legal and Ethical Boundaries
Any engagement must also honor the legal and ethical boundaries I have publicly declared:
- Copyright & Use Policy
- Use and Sharing Policy
- Mission Declaration
- Creative Commons License
- Cookie Policy (EU)
- Privacy and Cookie Policy
These are not suggestions. They are required terms of use.
This work did not come from funding. It came from persistence and personal cost.
Craig
Craig J. Phillips, MRC, BA
Individual living with the impact of a traumatic brain injury, Professional Rehabilitation Counselor, Author, Advocate, Keynote Speaker and Neuroplasticity Practitioner
Founder of Second Chance to Live
Founder of the Second Chance to Live Trauma-Informed Care AI Collaboration Model™
Leave a Reply