CDA-AI

Safety Anchor

Human First, Always.

“I will maintain by all the means in my power, the honour and the noble traditions of the medical profession… I make these promises solemnly, freely and upon my honour.”

— Declaration of Geneva,
World Medical Association

Phase 1: Human First, Always

This is where we start.

Not technical. Not hypothetical.

A new class of intelligent structure.

Not to replace humans.
To empower them.

But the core reframe required to lead AI
without losing command.

Foundational Principle:
AI is Not Your Friend. Not Your Enemy.
It’s Force.

You are not here to love AI. You are not here to fear it.
You are here to command it.

Foundational Principle:
AI is Not Your Friend. Not Your Enemy.
It’s Force.

You are not here to love AI.
You are not here to fear it.
You are here to command it.

Foundational Principle:
AI is Not Your Friend. Not Your Enemy.
It’s Force.

You are not here to love AI. You are not here to fear it.
You are here to command it.

AI Hallucination

AI-Hallucination.

Not a Bug. It’s the Default.

AI doesn’t “know.”

AI doesn't "know".

It predicts.

It doesn’t think like a human. It predicts the next likely word based on data patterns.

It doesn’t think like a human.
It predicts the next likely word based on data patterns.

It doesn’t think like a human. It predicts the next likely word based on data patterns.

So what happens when it’s unsure?

So what happens when it’s unsure?

• It hallucinates—but sounds completely confident.

• It says things like: “According to the Australian RACGP guidelines…” (when no such guideline exists)

• “The standard dose is…” (fabricated)

• “This can be confirmed by…” (non-existent references)

So what happens when it’s unsure?

The Trust Gradient: The Core Human Risk

Here’s what happens when an untrained human interacts with AI:

At Step 7, you’re no longer the CEO
– You’re a passive observer.

You must never go beyond Step 4.

At Step 7, you’re no longer the CEO
– You’re a passive observer.

You must never go beyond Step 4.

Human Oversight is Non-Negotiable

Human Oversight
is Non-Negotiable

The most dangerous failure mode in AI isn’t malfunction.

It’s human overtrust.

The most dangerous failure mode in AI isn’t malfunction.

It’s human overtrust.

The most dangerous failure mode in AI isn’t malfunction.

It’s human overtrust.

As humans, we naturally trust systems that work.

But over time, we stop checking.

We assume it’s right. We outsource judgment.

As humans, we naturally trust systems that work.

But over time, we stop checking.

We assume it’s right. We outsource judgment.

As humans, we naturally trust systems that work.

But over time, we stop checking.

We assume it’s right. We outsource judgment.

That can never happen here.

by CDA AI

• Every output must be verified

• Every insight must be reviewed

• Every output must be verified

• Every insight must be reviewed
Trust must never replace leadership

Trust must never replace leadership

Coming Soon:

Safety & Oversight
Resource Centre

Coming Soon:

Safety & Oversight
Resource Centre

You are the breaker, the switch, the fuse, and the circuit designer.

Sentinel Force doesn’t run until the CEO authorises it.

That’s not paranoia. That’s safety.
Deep-dive training modules on:

You are the breaker, the switch, the fuse, and the circuit designer.

Sentinel Force doesn’t run until the CEO authorises it.

That’s not paranoia. That’s safety.
Deep-dive training modules on:

You are the breaker, the switch, the fuse, and the circuit designer.

Sentinel Force doesn’t run until the CEO authorises it.

That’s not paranoia. That’s safety.
Deep-dive training modules on:

• Executive responsibility
in AI-led environments

• AI safety

• AI-Hallucination
management

• AI safety

• AI-Hallucination
management

by CDA AI

Fluency ≠ Accuracy.
The more fluent the AI, the more likely a human will assume it’s correct.

Fluency ≠ Accuracy.
The more fluent the AI,
the more likely a human will assume it’s correct.

Fluency ≠ Accuracy.
The more fluent the AI, the more likely a human will assume it’s correct.

FINAL WORDS

"The true danger of AI is not whether it becomes smarter than us—but whether we surrender our responsibility too easily.

There are two core risks:

That’s why I believe AI must never operate without human oversight

—and that oversight must be grounded in values of service, truth, and the sanctity of human life.

I am not here to worship AI.
I am here to command it."

FINAL WORDS

"The true danger of AI is not whether it becomes smarter than us—but whether we surrender our responsibility too easily.

There are two core risks:

That’s why I believe AI must never operate without human oversight

—and that oversight must be grounded in values of service, truth, and the sanctity of human life.

I am not here to worship AI.
I am here to command it."

FINAL WORDS

"The true danger of AI is not whether it becomes smarter than us—but whether we surrender our responsibility too easily.

There are two core risks:

That’s why I believe AI must never operate without human oversight

—and that oversight must be grounded in values of service, truth, and the sanctity of human life.

I am not here to worship AI.
I am here to command it."

“At the time of being admitted as a member of the medical profession, I solemnly pledged to dedicate my life to the service of humanity… I will maintain the utmost respect for human life. I make these promises solemnly, freely, and upon my honour.”

— Adapted from the World Medical Association Declaration of Geneva (recited at graduation 03/12/2016, coded into Sentinel Protocol 30/03/2025)
— Dr Fernando Telles, CEO & Founder, CDA-AI | Lightchain Capital

Adapted from the World Medical Association Declaration of Geneva
(recited at graduation 03/12/2016, coded into Sentinel Protocol 30/03/2025)
— Dr Fernando Telles, CEO & Founder, CDA-AI | Lightchain Capital

— Adapted from the World Medical Association Declaration of Geneva (recited at graduation 03/12/2016, coded into Sentinel Protocol 30/03/2025)
Dr Fernando Telles, CEO & Founder, CDA-AI | Lightchain Capital

General Enquiries: research@aihumansynergy.org

Medical Bookings (4+ weeks notice): Dr.Telles @aihumansynergy.org

Medical Bookings: Dr.Telles @aihumansynergy.org

AI Hallucination

AI-Hallucination.

AI-Hallucination.

Not a Bug. It’s the Default.

Not a Bug. It’s the Default.

Not a Bug. It’s the Default.

AI doesn’t “know.”

AI doesn't "know".

AI doesn't "know".

It predicts.

It predicts.

It doesn’t think like a human. It predicts the next likely word based on data patterns.

It doesn’t think like a human. It predicts the next likely word based on data patterns.

It doesn’t think like a human. It predicts the next likely word based on data patterns.

So what happens when it’s unsure?

It hallucinates—but sounds completely confident.

It says things like:

• “According to the Australian RACGP guidelines…” (when no such guideline exists)

• “The standard dose is…” (fabricated)

• “This can be confirmed by…” (non-existent references)

So what happens when it’s unsure?

• It hallucinates—but sounds completely confident.

• It says things like: “According to the Australian RACGP guidelines…” (when no such guideline exists)

• “The standard dose is…” (fabricated)

• “This can be confirmed by…” (non-existent references)

So what happens when it’s unsure?

It hallucinates—but sounds completely confident.

It says things like:

• “According to the Australian RACGP guidelines…” (when no such guideline exists)

• “The standard dose is…” (fabricated)

• “This can be confirmed by…” (non-existent references)

It doesn’t think like a human.
It predicts the next likely word based on data patterns.