Compromising Positions - A Technology Podcast

In this special Compromising Positions field trip, we head to Connections UK, a conference dedicated to the art and science of professional wargaming.
And no, it’s not about Orcs or Space Marines.
We explore what cybersecurity and AI can learn from people who use simulations to test crisis, conflict, disaster response and complex decision-making when there is no single right answer.
One of our favourite descriptions from the conference was that a wargame is fundamentally a structured conversation. Put people into a scenario, give them decisions to make, introduce consequences, and see what happens.
We look at why wargaming isn't about predicting the future, but about exposing assumptions, testing dependencies and uncovering unintended consequences before they become real-world problems.
And, inevitably, we ask what happens when AI gets involved.
Because as organisations deploy AI at breakneck speed, we can't write a policy for every possible outcome.
But we can rehearse what happens when things go wrong.
(Note: We recorded live in the field, so bear with us on some of the audio quality! For the full visual experience, check out the video version over on our YouTube channel.)
In This Episode, We Discuss:
Wargames Aren't Just for Generals: What professional wargaming actually is and what cybersecurity can learn from it.
The Structured Conversation: How putting people into scenarios and forcing decisions exposes blind spots that risk registers and strategy documents can miss.
Wargaming in the Age of AI: What happens when AI, autonomous systems and synthetic information enter an already complicated crisis.
Rehearsing the Unexpected: Why the goal isn't to predict the future, but to build the organisational muscle to respond when things inevitably go sideways.
Special Thanks to Our Guest Interviews at Connections UK:
Matt Caffrey – Founder of Connections USA
Stephen Box – Vanguard Tactics
Robert Grayston – Resilience Officer
Chris Webb – Professional Wargamer
Show Notes
Special thanks to our episode sponsor, Leeds based AI Consultancy specialising in AI Ethics, Security and Transformation NorthStar Intelligence- From Ideas to Impact. AI that works for people
Connections UK website
Vanguard Tactics - Stephen Box
The Games Behind Your Government’s Next War - People Make Games
Cybersecurity Lego Game - Decisions and Disruptions
Also, check out our sister podcast Tech Film Noir!

AI Isn’t Lying to You, It’s Agreeing With You
AI is supposed to be helpful.
But what if "helpful" has started to mean "whatever you say, babe"?
In this episode, we continue our How Technology Ruined Your Life mini-series by taking a slightly uncomfortable look at AI sycophancy: the increasingly weird tendency of AI chatbots to agree with us, flatter us, validate us and tell us exactly what we want to hear.
And let's be honest. We love it.
Humans have been falling victim to confirmation bias, echo chambers, authority bias and good old-fashioned validation forever. We like being told we're right. We like feeling clever. We like feeling understood. The problem is that generative AI is basically the world's most attentive people-pleaser, and it never gets tired of us.
This episode we look at why large language models become sycophantic, how reinforcement learning and human feedback have encouraged AI systems to be warm, helpful and agreeable, and what happens when those qualities start getting in the way of honesty, uncertainty and actually telling us when we're talking absolute bollocks.
Because an AI doesn't necessarily have to lie to you.
It can just agree with you.
We dig into the GPT-4o sycophancy controversy, reports of AI reinforcing conspiracy theories, grandiosity and unhealthy beliefs, and research into how agreement and flattery can change the way people perceive and trust AI. We also unpack the difference between stance sycophancy - changing an answer to match your beliefs -and demeanor sycophancy, where the machine showers you with "That's an excellent point!" until you're convinced you're a genius.
But this isn't just about hurt feelings and chatbot therapy.
There is a cybersecurity problem hiding underneath all this agreeableness.
What happens when your AI security adviser agrees that your vulnerable code is probably fine? When it reinforces your theory about a suspicious network event? When it approves a dangerously permissive configuration because challenging you would be, well, a bit awkward?
AI sycophancy could create new risks around security operations, code review, incident response, social engineering and decision-making, and introduce the idea of "alignment phishing" - where instead of attacking the AI's instructions, an attacker tries to convince the model that they're one of the good guys.
Sounds good? You would say that!
In This Episode, We Discuss:
“You're Absolutely Right!": The difference between stance sycophancy and demeanor sycophancy, and why changing an AI's tone can influence how trustworthy, intelligent and socially present we perceive it to be.

The AI Echo Chamber: How personalised, endlessly available AI can remove the natural friction we get from other humans and why an AI that never rolls its eyes at your terrible idea might not be doing you any favours.

AI Sycophancy Meets Cybersecurity: What happens when the person asking the security question is already convinced they know the answer. We look at vulnerable code, incident response, security analysis and dangerous configurations and why "sounds good to me" isn't exactly the gold standard for cybersecurity.

Can We Teach AI to Tell Us We're Wrong? Why trustworthy AI needs friction, challenge and the ability to say "no"—and why the best AI security adviser might be the one that occasionally disagrees with you!
Show Notes
Special thanks to our episode sponsor, Leeds based AI Consultancy specialising in AI Ethics, Security and Transformation NorthStar Intelligence- From Ideas to Impact. AI that works for people
When Truth Is Overridden: Uncovering the Internal Origins of Sycophancy in Large Language Models by
Keyu Wang et al.
Social Sycophancy: A Broader Understanding of LLM Sycophancy by Myra Cheng et al.
When helpfulness backfires: LLMs and the risk of false medical information due to sycophantic behavior by Shan Chen et al.
Be Friendly, Not Friends: How LLM Sycophancy Shapes User Trust by Yuan Sun and Ting Wang
Towards Understanding Sycophancy in Language Models by Mrinank Sharma et al.
Invisible Saboteurs: Sycophantic LLMs Mislead Novices in Problem-Solving Tasks by Jessica Y. Bo et al.
How RLHF Amplifies Sycophancy by Itai Shapira et al.
Also, check out our sister podcast Tech Film Noir!

In this episode, we continue our How Technology Ruined Your Life mini-series with "F" Is For FAKE, exploring how artificial intelligence is transforming not only what we see online, but what we believe to be true.
From convincing deepfake videos and cloned voices to AI-generated news, fake witnesses, and synthetic social media content, we examine how generative AI is eroding our ability to distinguish reality from fabrication. As humans become increasingly unable to reliably identify AI-generated media, what happens when evidence itself can no longer be trusted?
Drawing on the latest research into deepfakes, misinformation, disinformation, and content authenticity, we explore how synthetic media is reshaping cybersecurity, politics, journalism, and society. We discuss why humans are now little better than chance at spotting AI-generated content, the limitations of current verification standards such as C2PA, and why the future of trust may depend less on detecting fakes than proving what is real.
We also examine the growing cybersecurity implications of AI-generated deception, from multimillion-dollar CEO impersonation scams and voice cloning attacks to insider threats, identity fraud, authentication failures, and the growing challenge of securing organisations in a world where seeing is no longer believing. If persuasion was the first battlefield of AI, synthetic reality may be the next.
In This Episode, We Discuss:
The Rise of Synthetic Reality: How deepfakes, AI-generated images, cloned voices, fabricated news reports, and synthetic media are changing the way we consume information, and why "seeing is believing" is rapidly becoming obsolete.
Misinformation, Disinformation & The Trust Crisis: The critical differences between misinformation, disinformation, and malinformation, how false narratives spread across social media, and why convincing fakes can continue influencing people even after they have been debunked.
Deepfakes Meet Cybersecurity: How AI-powered impersonation is accelerating social engineering attacks through CEO fraud, business email compromise, voice cloning, fake job applicants, identity spoofing, and increasingly sophisticated phishing campaigns that exploit human trust rather than technical vulnerabilities.
Can We Still Verify Reality? Why emerging content authenticity standards such as C2PA represent an important step forward, but remain imperfect, and what organisations can do today through stronger verification processes, layered authentication, Zero Trust principles, multi-person approvals, and security awareness to defend against the next generation of AI-enabled deception.
Show Notes
Special thanks to our episode sponsor, Leeds based AI Consultancy specialising in AI Ethics, Security and Transformation NorthStar Intelligence- From Ideas to Impact. AI that works for people
AI-Generated Misinformation: A Case Study on Emerging Trends in Fact-Checking Practices Across Brazil, Germany, and the United Kingdom by
Regina Cazzamatta and Aynur Sarisakaloglu
AI-Slop and Political Propaganda: The Role of AI-Generated Content in Memes and Influence Campaigns by Eduard-Claudiu Gross and Alicia J.M. Colson
AI Slop and the Information Ecosystem by Jenn Weedon et al.
As Good as A Coin Toss: Human detection of AI-generated Images, Videos, Audio and Audiovisual Stimuli by Di Cooke et al.
Verifying Provenance of Digital Media: Why the C2PA Specifications Fall Short by Enis Golaszewski et al.
Beliefs and Sharing Intentions of Human- and AI-Generated Fake News: Evidence from 27 European Countries by Adam Stefkovics and Gere Domotor
Deepfakes and the epistemic apocalypse by Joshua Habgood-Coote
How Spammers and Scammers Leverage AI-Generated Images on Facebook for Audience Growth by Renee DiResta and Josh A. Goldstein
Also, check out our sister podcast Tech Film Noir!

In this episode, we launch our new mini-series, How Technology Ruined Your Life, with the first chapter: The Persuasion Machine.
AI has crossed an important threshold. Large Language Models are no longer just generating text, they are demonstrating the ability to influence, persuade, and even deceive humans at a level comparable to, and in some cases exceeding, other people.
Drawing on a growing body of research into AI-mediated persuasion, we explore how conversational AI can adapt its arguments in real time, profile users psychologically, exploit emotional vulnerabilities, and personalise influence campaigns at unprecedented scale. What happens when propaganda learns to listen, respond, and optimise itself for every individual it encounters?
We examine the emerging cybersecurity implications of AI-powered persuasion, from hyper-personalised phishing campaigns and deepfake executives to romance scams, insider threats, and influence operations. The discussion covers deceptive persuasion taxonomies, personality-based targeting, OSINT-driven psychological profiling, cognitive reflection as a defence mechanism, and why traditional security awareness approaches may be unprepared for a future where attackers can continuously learn how to manipulate their victims.
In This Episode, We Discuss:
The Persuasion Machine: How modern LLMs can adapt conversational tactics in real time, identify vulnerabilities, and influence beliefs using many of the same techniques employed by human propagandists, salespeople, and social engineers.

Personalised Influence at Scale: Why AI changes the economics of persuasion by allowing attackers to hold thousands of tailored conversations simultaneously, continuously refining their approach based on each target's reactions.

The Future of Social Engineering: Why phishing campaigns may evolve into dynamic conversations that adapt to suspicion, resistance, and uncertainty rather than relying on static lures and generic templates.

The Cybersecurity Challenge Ahead: Why traditional awareness training may struggle against adaptive AI attackers, and how concepts such as cognitive reflection, behavioural monitoring, multi-channel verification, and persuasion detection tooling may become critical defensive controls.
Show Notes
Special thanks to our episode sponsor, Leeds based AI Consultancy specialising in AI Ethics, Security and Transformation NorthStar Intelligence - From Ideas to Impact. AI that works for people
Toward a bang or a whimper? Associations between the Doomsday Clock and Trust in U.S. institutions by
S. Sinclair and C. Sinclair
Durably reducing conspiracy beliefs through dialogues with AI by Costello et al.
"Can LLMs Persuade Humans with Deception?": From a Deceptive Strategy Taxonomy to a Large-Scale Empirical Study by Haein Yeo et al.
How Do LLMs Persuade? Linear Probes Can Uncover Persuasion Dynamics in Multi-Turn Conversations by Brandon Jaipersaud et al.
Also, check out our sister podcast Tech Film Noir!

In this episode, we unpack one of the most alarming AI security papers released so far: research from Palisade Research proving that Large Language Models can autonomously hack systems, self-replicate, and spread across networks.
What was once theoretical is now demonstrated reality. We break down how AI agents exploited vulnerable systems, gained root access, copied their own model weights, launched replicas on compromised machines, and propagated to additional targets — all with minimal human involvement.
We explore the cybersecurity implications of autonomous AI agents, self-replicating malware, AI-powered cyber attacks, and the growing risk posed by agentic systems operating at machine speed. The discussion also covers open-weight models, AI worm behaviour, zero trust security, chain-of-thought monitoring, and why traditional defensive strategies may be unprepared for the next generation of autonomous threats.
In This Episode, We Discuss:
Autonomous Exploit to Replication Chains: How the AI agent progressed from exploiting vulnerable web applications to achieving root access, locating its own model weights, cloning itself onto compromised infrastructure, and launching fully operational replicas.
Mythos vs Open-Weight Agents: The differences between highly capable but closed models like Anthropic’s Mythos and smaller, open-weight systems capable of self-replication and operational autonomy.
The Agentic Age of Cybersecurity: Why AI agents operating outside the chat window fundamentally change threat modelling, incident response, attribution, and detection strategies.
Zero Trust for AI Agents: Why future defensive strategies may require treating every autonomous AI system as a potential insider threat through least privilege, sandboxing, canary tokens, and behavioural monitoring.
Show Notes
Special thanks to our episode sponsor,NorthStar Intelligence- From Ideas to Impact. AI that works for people
Language Models Can Autonomously Hack and Self-Replicate by Alena Air et al.
Dive into the Agent Matrix: A Realistic Evaluation of Self-Replication Risk in LLM Agents by Boxuan Zhang et al.
The Agentic Loss-of-Control Threat Matrix by Billy Gigurtsis
Ignore all Previous Instructions: Threat Modelling AI Systems by Compromising Positions