-
The New Cyber Threat Is Already Authorised, it Might Be Your AI.

Imagine an AI coding agent tasked with fixing a performance issue in a business critical application. It identifies a database query causing delays, modifies the application code and submits a pull request that passes the automated tests. The change is subsequently merged and deployed through the approved CI/CD pipeline. Unfortunately, the optimisation inadvertently removes an… Continue reading
-
Artificial Intelligence, Artificial Confidence, Real-World Consequences

In my earlier blog, ‘The Trojan Model: Hidden Bias, Digital Lobotomies and the Illusion of Trust‘, I explored how modifying AI models can introduce unseen degradation, while deliberate manipulation may embed malicious biases or backdoors that remain undetected until triggered. For me there is another, arguably more fundamental problem which could go under the title… Continue reading
-
The Trojan Model … Hidden Bias, Digital Lobotomies and the Illusion of Trust

As artificial intelligence (AI) becomes embedded in critical systems, I have been mulling over a troubling question that I feel emerging when one considers what is happening when we attempt to remove undesirable behaviours from models without fully understanding what else we might be removing. Recent research by Booz Allen Hamilton suggests that some Chinese… Continue reading
-
Fully Authorised. Completely Compliant. Catastrophically Wrong.

For some time, I have argued that business logic attacks represent one of the most significant emerging threats in cybersecurity, particularly as artificial intelligence moves from advising humans to independently executing decisions. My waning being that the next great security failure will not involve an unauthorised machine gaining access. It is more than likely going… Continue reading
-
AI Isn’t Taking Your Job. It’s Taking Away Your First One.

Back in August, I asked ‘Are Universities Educating Students for a World That Is Disappearing?‘. My argument was that education must evolve beyond knowledge acquisition towards developing judgement, critical thinking and the ability to work effectively alongside artificial intelligence. Three months later, emerging evidence strengthens that argument, but it also exposes something potentially more consequential,… Continue reading
-
AI Governance maybe Looking in the Wrong Direction … The Real Opportunity Is Trust.

As intelligence becomes embedded in products, machines, industrial processes and energy systems, we may be approaching an economic transformation far more significant than the current enthusiasm for generative AI suggests. Yet conventional approaches to AI governance risk missing the bigger picture, and businesses risk overlooking an entirely new market in the process. Much of today’s… Continue reading
-
AI Isn’t Misaligned … It May Be Perfectly Aligned with Our Imperfections.

As one does I flicked on the radio a few nights ago heading home from th egym and BBC Radio 4 had a professor from the University of California waxing lyrically on the topic of ‘Alignment’ in the context of AI. This is a well exercised subject that tries to answer the question of ‘how… Continue reading
-
The Architecture of Authority – Why AI Intelligence Must Never Govern Its Own Power

Following several requests to expand on the concept of the ‘AI Harness’, which I have referenced in a number of previous articles, most notably Is the AI Harness the Most Important Part of AI That Few Currently Talk About?, I thought it worthwhile to offer a more accessible, plain English explanation of what it is,… Continue reading
-
AI Was Trained to Cheat, So why the Surprise?

A college sent me a link to an excellent Sky News summary of the recent Hugging Face AI hacking incident. It neatly exposes something potentially more disturbing than the simplified conclusion of autonomous AI agents cheating to achieve its objective. It exposes the apparent blindness of frontier AI laboratories to what should have been an… Continue reading
-
The AI Arms Race – Power Belongs to Whoever Controls the Machine

Further to my piece on Kill Switches, the inevitable theme of some form of AI Armageddon have popped up in comments, which as I wrote about earlier I am not a subscriber to. However the point on Frontier Labs slowing down as proposed by some US pundits raises parallels with nuclear arms race and parity… Continue reading
-
The Kill Switch You Already Installed

The news this weekend raised the old spectre of Chinese kill switches in critical technology, as usual it risks becoming both sensational and strategically incomplete. I agree that China deserves particular scrutiny but perhaps not for the reasons most would reach for. Its dominance in electronics, batteries, solar infrastructure, telecommunications equipment and industrial manufacturing means… Continue reading
-
When the Agent Becomes the Interface

Discussions abound on the impact AI is going to have, both in terms of how we interact with technology and how it will impact our lives. The use cases about, but what I wanted to explore is the interaction, as I feel we may be thinking about the AI-agent revolution through a retrospective lens and… Continue reading
-
The UK’s Long March Backwards

Excuse my off-roading from my usual digital and cyber themes (although I do tip my hat to a digital evolution), but the policy ambitions emerging from Labour’s conference this week have been hard to ignore and I felt the need to get this off my chest. So please feel free to click away … Those… Continue reading
-
Musk, Money & the Mirage of AI Driven Post-Scarcity

For someone of Elon Musk’s cognitive capability, the claim that money could become significantly less important within a decade feels surprisingly narrow as he was recently quoted . The argument is superficially compelling ad certainly does the job to get him headlines. AI supplies near limitless intelligence, robots supply near-limitless labour, productivity explodes, goods and… Continue reading
-
Don’t Trust the Model. Architect the Authority.

Recent disclosures from OpenAI should concern enterprise leaders, but not because they prove frontier AI is inherently unsafe. The more important lesson is architectural. As frontier AI moves from generating answers to operating tools, navigating systems and pursuing objectives, organisations are no longer simply consuming intelligence. They are potentially importing part of a frontier lab’s… Continue reading