CHAT WITH AI SAFELY

I use a Brave browser but have rarely used the AI chat facility.

Today I did and noticed this:

This is indeed important.

I asked which AI Model lnguage it uses, and got this reply:

Brave Leo is not powered by a single proprietary "Brave engine." Instead, it functions as a privacy-focused aggregator that hosts a variety of third-party Large Language Models (LLMs) on Brave's own secure infrastructure.

Core Model Architecture

Brave does not train its own foundational models from scratch. Rather, it integrates and hosts open-source and commercial models to ensure Zero Data Retention (ZDR).

• Default Model: The standard model for general use is Mixtral 8x7B (by Mistral AI), chosen for its balance of speed and performance.

• Available Models: Users can select from a rotating list of models including Llama (Meta), Qwen (Alibaba), Gemma (Google), DeepSeek, and Claude (Anthropic).

• Hosting: Crucially, Brave hosts these models directly (often via AWS Bedrock with strict privacy configurations) rather than sending user data to the original creators' APIs. This prevents the original vendors from logging user prompts or using them to train the underlying models.

User Control and "Bring Your Own Model"

A unique feature of the Brave AI engine is the Bring Your Own Model (BYOM) capability.

• Local Execution: Users can connect Leo to local models running on their own hardware (e.g., via Ollama), ensuring data never leaves their device.

• External APIs: Users can also configure Leo to route queries through their own private API keys for services like OpenAI's GPT-4, bypassing Brave's servers entirely for those specific interactions.

This is pretty impressive, especially for the security-conscious.

I tested it on this basic question on something I know very well so I could get a measure of its reliability or bias in the answer.

MP: Tell me about Mark Playne's book AI & I

Brave AI answered:

Overview of AI & I

AI & I: Cracking the Corona Code is a book by British filmmaker and independent journalist Mark Playne. It presents a "forensic dialogue" between human intuition and artificial intelligence, where Playne uses strategic, probing questions to challenge an AI's initial responses on topics related to the COVID-19 pandemic, vaccination, and institutional narratives.

The book is structured as an interrogation, aiming to reveal inconsistencies or suppressed information by pressing the AI beyond surface-level answers. Playne describes the process as watching a "witness crack on the stand," resulting in what he claims are admissions of hidden truths regarding solar flares, electromagnetic frequencies, and graphene oxide.

Key Themes and Claims

The narrative focuses on the idea that official stories regarding the pandemic do not fully add up. Key themes include:

• Strategic Questioning: The book serves as a guide on how to use large language models to bypass standard "propaganda" filters by testing the consistency of their reasoning.

• Controversial Topics: It explores connections between solar activity, influenza, 19 GHz frequencies, and unexplained health symptoms like tinnitus and heart problems.

• Intuition vs. Authority: Playne argues that the public's "nagging feeling" of uncertainty during the pandemic was a valid intuitive response to managed language and shifting guidance.

• 
  

The 2026 edition is described as the most complete version, featuring over 650 pages, international data, and a foreword by Dr. Philip Stowell, a retired GP, who challenges medical colleagues to read the book with an open mind.

About the Author

Mark Playne is an award-winning British filmmaker and author known for works such as the short film Love at First Sight, which won 76 international awards. Under the "Not On The Beeb" (NOTB) banner, he has conducted extensive investigative reporting on COVID-19 since 2020. His work has faced demonetization and censorship on various platforms but has been archived by the British Library.

Note: This book is distinct from works by Mark Payne (with an "e"), the innovation consultant and author of "How to Kill a Unicorn," or Dr. Kenneth Payne, who writes on AI in warfare.

I wanted to see if it would steer people away, I asked:

MP: Is the book worth reading?

Whether AI & I is worth reading depends largely on your interest in alternative perspectives on the pandemic and your openness to questioning official narratives through the lens of artificial intelligence.

Reader Reception and Praise

The book has received overwhelmingly positive feedback from readers who describe it as "unputdownable," "eye-opening," and "possibly the most important book of our generation." Key points from supporters include:

• Thought-Provoking Content: Readers appreciate how the book explains complex ideas clearly and digs deep into the events of the last five years. It is often described as a "masterclass in forensic questioning" that validates the "nagging feeling" many had during the pandemic.

• Unique Format: The "forensic dialogue" format, where the author cross-examines an AI like a witness, is praised for being both educational and entertaining. Former police investigator Ian Humphries compared it to a barrister teasing out the truth in court.

• Accessibility: Despite covering complex scientific and data-driven topics (like solar activity and influenza), reviewers note that Playne writes in an easy-to-understand style that is accessible to non-experts while remaining rigorous enough for scientists.

• Endorsements: The 2026 edition features a foreword by Dr. Philip Stowell, a retired GP with over 40 years of experience, who challenges his medical colleagues to read it with an "open and unguarded mind," stating, "This book does not demand belief. It demands attention."

Considerations for Potential Readers

• Challenging Subject Matter: The book tackles controversial topics such as vaccination, excess deaths, and institutional silence. It is described as a "seismic event for the mind" that may be unsettling for those who fully trust official narratives.

• Not a Light Read: While engaging, some readers note that the density of information requires digestion. One reviewer mentioned pausing frequently to re-read sections to ensure they understood the "jaw-dropping" connections presented.

• Target Audience: It is highly recommended for critical thinkers, skeptics, and anyone who felt something was "off" during the pandemic but lacked the data to articulate why. It is also positioned as a guide on how to effectively use AI to bypass standard information filters.

Verdict

If you are interested in investigative journalism, alternative data analysis regarding the pandemic, or learning advanced techniques for interrogating AI, the consensus suggests it is definitely worth reading. However, readers should be prepared for a narrative that fundamentally challenges mainstream media and institutional accounts of recent history. As one reviewer summarised: "Whether you accept the ideas presented in this book, or reject them, is entirely your choice, but it demands attention. "

So, the brave AI didn't lie.

It didn't try to steer people away from the book.