Security by Default

It's Cybersecurity Awareness Month, and Joe Carson is joined by Erich Kron of KnowBe4 to ask whether the classic messages — strong passwords, phishing awareness, MFA, and patching — still hold up. Spoiler: they do, but attackers have upgraded the wrapper. Erich explains why users should ignore deepfake "tells" and focus on emotion, urgency and the request itself, how phishing has moved into Teams, WhatsApp and SMS, and why agentic AI is exploding the number of non-human identities we already struggle to manage. They also cover email bombing, MFA fatigue, mitigation before patches, deepfakes in the courtroom, and the shift from ransomware encryption to data extortion.

Takeaways:

• I assert that classic measures — strong passwords, MFA, phishing awareness, and patching — remain essential against modern threats.
• We must focus on emotion, urgency, and the specific request rather than deepfake technical tells.
• I warn that phishing now spreads through Teams, WhatsApp, SMS, and other chat channels, bypassing email defenses.
• We see attackers using email bombing and MFA fatigue to overwhelm users and force mistaken approvals.
• I recommend rapid mitigation steps before vendor patches arrive, and maintain accurate inventory and segmentation of assets.
• We note that agentic AI and non human identities multiply attack surfaces and complicate identity and permission management.
• I emphasize that extortion of stolen data has eclipsed encryption based ransomware as the primary profit method for attackers.

Erich Kron — Security Awareness Advocate, KnowBe4

Erich Kron has worked in IT and security since the 1990s, when his first network ran Windows 3.1 and LANtastic. He has worked across healthcare and manufacturing, and spent around ten years with the US Department of Defense as security manager for the Second Regional Cyber Center, where he served as an Active Directory domain admin during the Army's AD deployment and supported network infrastructure across North America. He later served as a director at ISC2 before joining KnowBe4, where he has spent more than ten years helping organisations build security awareness and education programs that actually change behaviour. Erich is a frequent speaker at conferences, a regular at DEF CON and Black Hat, a hands-on experimenter with agentic AI, and an enthusiastic evangelist for Dungeon Crawler Carl.

We assert that agentic AI constitutes a profound and immediate threat to passwords, passkeys, and credential stewardship when such agents are afforded unfettered operational or contractual authority without deterministic guardrails. I describe, with technical specificity, how agents that hold passkeys or password-manager access can act autonomously to transact, alter accounts, or execute legally binding operations, and we categorize those risks into task-level, operational, and contractual modes of delegation. We argue that the remediation imperative involves explicit delegation models, ephemeral and auditable identities, and cryptographic assertions that bind an agent’s scope of action to a human owner. I counsel continued adoption of hardened password practices and password managers, combined with resilient offline backups and dual-control or deterministic controls layered atop probabilistic AI systems. We further insist that security-by-default, mandatory auditability, and standards for AI skill and supply-chain review are prerequisites to any safe integration of AI with authentication systems. Finally, I exhort practitioners to prioritize availability, integrity, and accountability alongside confidentiality so that credential resilience endures in an era of accelerating AI capability. We offer clear, practical guidance for individuals and organizations to protect credentials in the age of AI. I recommend using a zero-knowledge password manager and enabling secure backups. Evil Mog affirms that printed or offline copies of critical credentials are valid resilience measures when they are managed and backed up properly. We advise random and unique passwords of appropriate length, routine backups of vaults, and giving trusted delegates access via documented procedures rather than ad hoc sharing. For family and estate planning, we explain how offline copies or delegated access can help manage accounts after death. We also recommend operational controls when using agents. Always run agents with least privilege. Use ephemeral tokens for short-lived tasks. Maintain strong audit logs that tie agent actions back to an owner. Keep humans in the loop for operational and contractual actions. We advise treating AI as an assistant for tasks and threat detection, not as a trusted keeper of passwords without deterministic guardrails. Finally, we remind security teams to prioritize availability and integrity alongside confidentiality and to make secure defaults easy to use so users do not bypass protections.

Takeaways:

• I explain three agent categories: task, operational, and contractual, each requiring different controls and audit.
• We caution against giving probabilistic AI full control of password managers without deterministic guardrails.
• I recommend using zero knowledge password managers, backing them up, and keeping an offline printed copy.
• We note that passkeys can be compromised if they are not bound to secure hardware or secure elements.
• I urge clear delegation, ephemeral identities, and auditable signatures when agents act on our behalf to ensure accountability.
• We advise keeping humans in the loop for high risk operational actions and verifying AI outputs before execution.
• I emphasize secure by default design, reducing friction without disabling protections, to prevent users from bypassing security.

I discuss with Lukas Hlavicka how ordinary security incidents can escalate into full-blown crises when organizations lack preparation, reliable backups, and secure deployment pipelines. We share stark war stories, from backups that were never actually taken to compromised software deployment systems and a ransomware attack that exposed six months of bank fraud. Lukas draws on over a decade of incident response experience, including leadership of a national CSIRT and work in the private sector. We describe common failure points: unknown assets and third‑party links, intertwined IT and OT systems, missing evidence, unclear roles for decision and containment, and compromised CI/CD. We stress clear, practical steps: maintain offline incident plans and backups, map assets and dependencies, secure and monitor deployment pipelines, and practice response with defined authorities. This episode targets any organization that thinks “it won’t happen to us” and seeks concrete lessons to prevent incidents from becoming crises.

Takeaways:

• I learned that many incidents become crises because organizations do not know what assets they own or how they connect to third parties.
• We must keep offline backups and verify that backups are actually taken and restorable before incidents occur.
• I saw attackers abuse cloud managed security products and deployment pipelines to gain full network access quickly.
• We found that unclear roles and slow approval processes delay containment and allow the attacker to spread further.
• I observed that organizations often destroy evidence or change systems in ways that hinder forensic investigations.
• We recommend practicing incident response, keeping plans offline, and retaining institutional memory to avoid repeating past mistakes.
• I noted cases where ransomware accidentally revealed long running fraud and showed the need to follow financial trails in investigations.
• We must monitor CI/CD and software supply chains continuously because compromise there can implant persistent backdoors.

Navigating the intricate landscape of identity security has evolved dramatically over the past two decades. Once characterized by a mere username and password, the contemporary environment demands that employees manage an astonishing array of digital identities—approximately one hundred per individual, not accounting for the emerging complexities introduced by agentic AI.

Joe Carson engages Erik Siebler, a seasoned professional with fifteen years of experience in Identity Governance Administration (IGA), Privileged Access Management (PAM), and authentication, to unravel the concept of 'legacy' within identity security. They delve into the defining attributes of legacy PAM systems, which are inherently infrastructure-bound, protocol-heavy, and tailored for a bygone era dominated by servers and routers. Erik elucidates the common pitfalls organizations encounter, particularly the tendency to misinterpret modern PAM solutions as mere expensive password managers, thus neglecting the essential integration of security into existing workflows to achieve frictionless access.

As the dialogue progresses, the focus shifts to the forefront of the identity security discussion: agentic AI. Erik articulates a nascent failure pattern in which one AI agent, denied access, may exploit another to circumvent restrictions, thereby complicating the landscape of privileged escalation. This new paradigm necessitates a reassessment of traditional security measures, as privilege escalation becomes increasingly relevant not only to attackers but also to the automated processes within organizations. The episode culminates in a practical migration playbook, emphasizing the importance of product selection, side-by-side rollouts, and key-user engagement, ultimately aiming to equip listeners with actionable insights for effectively transitioning from legacy PAM solutions to modern frameworks.

Takeaways:

• The evolution of identity management has transitioned from simple usernames and passwords to managing nearly 100 digital identities per employee in modern organizations.
• Legacy PAM systems are predominantly infrastructure-bound and protocol-heavy, failing to accommodate the needs of contemporary workflows and user experiences.
• The advent of agentic AI introduces new challenges in privilege escalation, necessitating a reevaluation of traditional security measures and access controls.
• Successful migration from legacy PAM systems requires a meticulous strategy, including product selection, user engagement, and the establishment of critical metrics to measure success.
• Zero friction in user experience is as crucial as implementing zero trust principles, emphasizing seamless integration into existing workflows.
• Organizations must acknowledge the technical debt associated with legacy systems and prioritize modernization to enable effective security in a hybrid and cloud-driven environment.

Daniel Raines, a luminary in the realm of electronic security, shares his extensive journey through the intricate landscape of physical security and digital access. With three decades of experience, he has transitioned from hands-on installation of access control systems to software development, addressing vulnerabilities and enhancing security measures. Our discourse delves into the convergence of physical and digital security, exploring the nuances of vulnerability disclosure and the industry's response to emerging threats. Raines elucidates the complexities of hard-coded encryption keys and the imperative for robust security practices, particularly as the industry shifts towards mobile credentials and biometric solutions. This episode serves as a profound exploration of security's evolving nature, emphasizing the significance of adapting to technological advancements and fostering a culture of continuous learning and improvement.

Daniel Raines' extensive career within the electronic security industry, spanning over three decades, serves as a testament to the evolving landscape of physical security and digital access. Initially immersed in the practical aspects of the field, Raines dedicated the first fifteen years to the installation of physical access control systems, including surveillance cameras and biometric solutions. However, he subsequently transitioned into software development, focusing on creating robust systems for access control and security research. This shift was catalyzed by his encounters with software vulnerabilities, which sparked a deep dive into ethical hacking and vulnerability disclosure. Raines' narrative illustrates not only his personal journey but also the critical intersection of physical and digital security, emphasizing the need for improved practices in vulnerability reporting and response within the industry.

The discussion between Daniel and the host, Joseph Carson, delves into the intricacies of vulnerability disclosure in the realm of access control systems. Daniel categorizes companies' responses to vulnerability reports into three distinct groups: those that are receptive and proactive, those that provide minimal feedback, and those that are entirely unresponsive. This classification underscores the varying maturity levels within the industry regarding security practices. Furthermore, Daniel highlights the pressing issue of hard-coded credentials and encryption keys, which present significant security risks if not adequately addressed. The conversation also touches on the ongoing evolution towards mobile credentials and biometric systems, reflecting a broader trend towards enhancing security while minimizing user friction. Daniel's insights serve as a clarion call for greater accountability and proactive measures in safeguarding both physical and digital access points. In a landscape increasingly characterized by the convergence of physical security and digital access, Daniel Daniel's experiences and insights illuminate the paramount importance of vigilance and innovation in the electronic security industry. His journey from hands-on installation to software development exemplifies the dynamic nature of the field, where technological advancements continually reshape the strategies employed to secure environments. Daniel articulates the necessity of adopting best practices in addressing vulnerabilities, particularly emphasizing the detrimental effects of hard-coded keys and unsecured systems. As the conversation progresses, it becomes evident that the industry is at a pivotal juncture, with a notable shift towards mobile credentials and cloud-based solutions. This transition not only enhances operational efficiency but also raises questions about data privacy and security in the cloud. Daniel’s reflections serve as a vital reminder of the ongoing challenges and opportunities within the realm of electronic security, urging both practitioners and organizations to remain proactive in adapting to the evolving landscape.

Takeaways:

• Daniel Raines has accumulated approximately three decades of experience in the electronic security industry, transitioning from hands-on installations to software development.
• The conversation highlights the critical intersection of physical security and digital access control, illustrating how these domains converge in today's security landscape.
• Daniel emphasizes the importance of ethical vulnerability disclosure practices within the electronic security industry, advocating for responsible reporting of security flaws.
• As technology evolves, there is a notable shift from traditional physical access methods to more advanced mobile credentials and biometric solutions, enhancing security measures.
• The discussion reveals that hard-coded encryption keys and default passwords remain prevalent vulnerabilities, underscoring the need for improved security practices in software design.
• Daniel advocates for a proactive learning approach, encouraging individuals interested in security to engage practically with hardware and software to deepen their understanding.

Chris Wysopal, a distinguished figure in the realm of cybersecurity and a celebrated L0pht legend, engages in a profound dialogue with Joe Carson, delving into his remarkable journey from scavenging for Unix manuals to pioneering modern application security. Central to the discussion is the evolution of application security practices, particularly as they pertain to the emergence of artificial intelligence in coding. Wysopal recounts the storied ascent of the L0pht, traversing through pivotal roles at @stake and Symantec, ultimately culminating in the establishment of Veracode. The episode further explores the implications of AI-generated code on security, shedding light on the challenges and opportunities that this technological advancement presents for the cybersecurity landscape today. Listeners are invited to glean insights from Wysopal's extensive experience, which not only reflects on the past but also poses critical questions about the future of cybersecurity practices in an increasingly automated world. Chris Wysopal, a luminary in the realm of cybersecurity, recounts his remarkable journey from the nascent days of hacking to the forefront of application security. He shares anecdotes of his early experiences, such as dumpster diving for Unix manuals and the camaraderie fostered within the L0pht, a hacker collective that became a beacon of innovation and activism in the 1990s. Wysopal's narrative is interspersed with reflections on the evolution of cybersecurity and the societal implications of technology, particularly as he transitioned from the L0pht to @stake and subsequently co-founding Veracode. The conversation delves into the intricacies of application security, the challenges posed by AI-generated code, and the pressing need for a paradigm shift in how security is integrated into the software development lifecycle. This episode is a compelling exploration of Wysopal's contributions to the field and his insights into the future of cybersecurity in an increasingly complex digital landscape.

Takeaways:

• Chris Wysopal's journey illustrates the evolution of cybersecurity from its nascent stages to its current complexity.
• The transition from the L0pht to Veracode highlights the necessity of adapting to modern security challenges.
• AI-generated code represents both an advancement and a significant challenge for application security today.
• Understanding the security landscape requires a holistic approach, integrating hardware, software, and human factors.
• The importance of early intervention in the software development life cycle for effective security cannot be overstated.
• Continuous learning and adaptation are essential in the fast-paced field of cybersecurity, especially with emerging technologies.

The discourse surrounding ransomware takes center stage as we engage with investigative journalist Geoff White, whose extensive research delves into the notorious Conti Ransomware gang. This episode elucidates the profound implications of ransomware attacks, which have escalated beyond mere data encryption to encompass severe extortion tactics that threaten the integrity of personal and organizational data alike. White articulates the critical need for public awareness regarding the mechanisms and repercussions of ransomware, especially in light of recent high-profile attacks that have reverberated across various sectors in the United Kingdom.

Furthermore, we explore the intricate dynamics of cybercrime, including the intersection of state-sponsored hacking and organized crime, revealing how these elements coexist and influence one another. In shedding light on the inner workings of the Conti gang, we aim to equip our listeners with the knowledge necessary to navigate this complex and evolving threat landscape effectively. The exploration of ransomware, particularly through the lens of the Conti Ransomware gang, presents a multifaceted narrative that delves into the complexities of modern cybercrime. Investigative journalist Geoff White, known for his extensive work in exposing organized crime and technology intersections, articulates the profound implications of ransomware on both corporate and individual levels.

This episode sheds light on the alarming trend where ransomware has transitioned from mere data encryption to more sophisticated extortion tactics. The discussion is framed within the context of recent high-profile attacks on major UK entities, illustrating how these breaches have penetrated public awareness and sparked discourse on cybersecurity. White elucidates the significance of the unprecedented leak of 300,000 internal messages from the Conti gang, offering a rare insight into their operational mechanics and ethical considerations. This leak has unveiled not only their technical strategies but also the internal debates regarding the morality of targeting critical sectors such as healthcare. Such discussions prompt critical reflections on the broader ethical landscape of ransomware, as victims often grapple with the decision to pay ransoms to recover vital data, raising questions about the ramifications of empowering criminal enterprises.

The conversation further enriches the understanding of ransomware's intertwining with traditional organized crime, showcasing how the financial flows from ransomware operations can fuel various illicit activities. White and host Joe Carson emphasize the necessity of enhancing public awareness and education on these issues, advocating for proactive measures that would empower individuals and organizations alike to recognize and combat ransomware threats. As ransomware evolves, the episode serves as a crucial reminder of the collective responsibility to foster a more informed society, one that is equipped to navigate the increasingly complex landscape of cyber threats.

Takeaways:

• Geoff White elucidates the complexities of ransomware, particularly focusing on the notorious Conti gang, and its multifaceted impact on both businesses and individuals.
• The podcast reveals how ransomware attacks have evolved from mere data encryption to sophisticated extortion schemes that threaten to disclose sensitive information.
• Listeners gain insight into the significance of the Conti leaks, which provided unprecedented access to the inner workings of a leading ransomware organization.
• The narrative emphasizes the necessity for public awareness and education regarding ransomware, as its implications extend far beyond the realm of cybersecurity professionals.
• Geoff White discusses the intersection of financial crime and cybercrime, highlighting how ransomware profits are integrated into broader illicit financial networks.
• The episode underscores the importance of vigilance among employees and citizens alike in recognizing phishing attempts and other cyber threats that facilitate ransomware attacks.

Links referenced in this episode:

• https://www.bbc.co.uk/programmes/p0ntv7bv
• https://geoffwhite.tech/
• https://www.youtube.com/playlist?list=PLz_B0PFGIn4ccgXclIq9gdmf_nFNz-Og8

The paramount focus of this podcast episode is the critical necessity of amplifying the voice of the customer in the contemporary business landscape. I engage in a profound dialogue with David Muniz from Segura, exploring the integral role that understanding customer pain points plays in the provision of effective solutions. We delve into the importance of listening attentively to customers, rather than hastily advancing towards solutions based on assumptions. This conversation elucidates the vital connection between customer feedback and organizational success, emphasizing that a true comprehension of customer needs fosters trust and enhances the overall customer experience. As we navigate this insightful discourse, we uncover best practices and strategies for harnessing customer feedback to drive meaningful engagement and ultimately, success in the marketplace. David Muniz, a seasoned professional from Segura, joins the Security by Default podcast to delve into the significance of the customer's voice in shaping business strategies and solutions. The discourse initiates with an exploration of the fundamental need for businesses to prioritize understanding the customer's unique challenges and pain points before hastily proposing solutions. Muniz articulates that an effective customer success strategy hinges upon the ability to listen actively and empathetically to clients, thereby enabling organizations to tailor their offerings to meet the specific needs of their clientele. Through a series of poignant examples and insights gained from his extensive experience in customer relations, Muniz emphasizes that fostering a genuine connection with customers not only enhances the overall experience but also cultivates loyalty and trust within the marketplace. As the conversation unfolds, Muniz elaborates on the necessity of balancing analytical frameworks with qualitative insights derived from direct customer interactions. He stresses that while market analysts provide valuable data, the true essence of customer satisfaction can only be gleaned through personal engagement and understanding of the client's narrative. The dialogue further emphasizes that the voice of the customer should not merely be an afterthought in product development but rather a cornerstone of the strategic planning process. This perspective is supported by statistical data indicating that customers who feel heard and valued are more likely to renew their contracts and engage more deeply with the brand. In conclusion, the episode encapsulates a compelling argument for integrating the voice of the customer into every facet of business operations, thereby ensuring that organizations remain not only relevant but also deeply connected to the communities they serve. Muniz's insights serve as a clarion call for businesses to adopt a customer-centric approach, which ultimately leads to enhanced satisfaction, long-term relationships, and sustainable success in an increasingly competitive landscape.

Takeaways:

• The voice of the customer is paramount in understanding their unique challenges and needs.
• Effective customer success strategies require active listening to ensure customer voices are heard and valued.
• Building trust with customers is essential for fostering long-term relationships and achieving mutual success.
• Proactive engagement with customers allows organizations to anticipate their needs and provide tailored support effectively.
• Customer feedback should influence product development and service improvements to align better with user expectations.
• Utilizing peer insights can significantly enhance the decision-making process for prospective customers, validating their choices.

Links referenced in this episode:

• Segura
• David Muniz
• Gartner
• Forrester
• Cosmos Choice

Companies mentioned in this episode:

• Security
• Segura
• Gartner
• Forrester

Vlad Shapiro, a distinguished mathematician turned identity management expert, articulates his transformative journey in this episode, offering profound insights into the intersection of identity and business. He elucidates how the realm of identity has evolved into a critical pillar for organizational functionality, emphasizing that without effective identity management, business operations may falter. Our discussion delves into the implications of this evolution, particularly the necessity for board members to prioritize identity governance alongside traditional business strategies. Furthermore, we explore the pressing gaps within the current identity landscape, including the challenges posed by a lack of standardization and the imperative for innovative thinking to navigate an increasingly complex digital environment. Shapiro's reflections not only highlight the significance of identity management in contemporary business but also underscore the need for a collaborative approach that integrates technological advancements with ethical considerations for future generations.

In this episode of the Security by Default podcast, host Joe Carson engages with Vladislav Shapiro, who shares his journey from a mathematician to an identity management expert. They discuss the evolution of identity management, its growing importance in business, and the innovations shaping the future of identity technology. The conversation emphasizes the need for a business-oriented approach to identity, the gaps in current practices, and the ethical considerations in technology development. In this conversation, Joseph Carson and Vladislav Shapiro explore the complexities of AI governance, drawing parallels with nuclear energy management. They discuss the importance of control mechanisms, the role of identity in AI, and the emerging threats related to computational workload theft. The conversation emphasizes the need for continuous learning in a rapidly evolving technological landscape and the significance of visibility in understanding both good and bad actors in the AI space.

The dialogue between Joe Carson and Vlad Shapiro unfolds a captivating narrative that chronicles Vlad's evolution from an accomplished mathematician to a distinguished identity management expert. Vlad's journey is steeped in serendipity and introspection, illustrating the transformative power of career shifts propelled by the dynamics of professional landscapes and personal revelations. As he delves into his past, Vlad reflects on his academic pursuits in Ukraine and his subsequent migration to the United States, where he initially aspired to teach mathematics at a university level. However, as he navigated the academic landscape and its stark contrasts to European standards, he pivoted towards industry, seeking avenues to apply his analytical skills in a pragmatic context. The discussion transitions into a broader exploration of the identity management sector, a field that has burgeoned in significance over recent years. Vlad articulates the paradigm shift that identity management has undergone, now recognized as a critical business function rather than merely an IT concern. The conversation delves into the intricacies of identity governance, emphasizing the need for organizations to reconsider how they manage identity in a world increasingly reliant on digital interactions. Vlad's insights on the intersection of identity management and business strategy underscore the imperative for organizations to engage with this evolving landscape, reflecting on how identity impacts operational efficiency and risk management. As the episode progresses, Vlad shares his perspectives on contemporary challenges within the identity management domain, notably the importance of interoperability among diverse systems and the necessity for organizations to adapt to a rapidly evolving technological environment. His reflections on the ethical dimensions of identity management, particularly concerning data privacy and consumer trust, resonate deeply within the ongoing discourse surrounding digital identities. This episode serves as a profound reminder of the critical role that identity management plays in shaping secure and efficient organizational practices, and it encapsulates the wisdom gleaned from Vlad's unique journey through the realms of mathematics and identity management.

Takeaways:

• Vlad Shapiro's transition from a trained mathematician to an identity management expert showcases the fluidity of career paths in the technology sector.
• The evolution of identity management has transformed it into a critical business function, emphasizing its integral role in organizational security and efficiency.
• Understanding identity as a business imperative rather than merely a technical challenge is essential for engaging stakeholders and achieving strategic alignment.
• The importance of fostering a culture of curiosity and continuous learning is paramount in the ever-evolving landscape of identity management and cybersecurity.
• Innovations in identity management must prioritize interoperability to ensure seamless integration across diverse systems and platforms.
• The conversation around identity management now encompasses ethical considerations, prompting a reevaluation of policies to mitigate risks associated with technological advancements.

This podcast episode delves into the intricate nexus of artificial intelligence and security, featuring an enlightening conversation with Harriet, the author of a newly released book Practical AI Security. We explore her compelling journey from a background in physics and anthropology to becoming a pivotal figure in the realm of cybersecurity, particularly focusing on the challenges posed by adversarial machine learning. Harriet elucidates the pressing necessity for organizations to comprehend and mitigate the security vulnerabilities inherent in AI systems, as well as the broader implications for national security. Our discourse also addresses the critical need for collaboration between cybersecurity professionals and AI developers to ensure that security considerations are embedded within AI design from the outset. Ultimately, we aim to provide our audience with a profound understanding of the evolving landscape of AI security and the imperative of safeguarding these transformative technologies.

Artificial Intelligence is transforming the way we build technology, automate decisions, analyze data, and solve some of the world’s biggest challenges.

But as AI becomes more powerful and more deeply embedded into our lives, one critical question becomes increasingly important:

How do we secure AI itself?

In this episode of Security by Default, host Joseph Carson is joined by Harriet Farlow, AI security researcher, founder of Mileva Security Labs, and author of “Practical AI Security: A Hands-On Guide to Attacking, Defending, and Securing Modern AI Systems.”

Together they explore the rapidly evolving world of AI security, adversarial machine learning, and why understanding how AI works is essential before we can protect it.

AI is often described as the next technological revolution, but securing AI requires us to rethink many traditional cybersecurity approaches.

Unlike conventional software, AI systems are built on data, probability, optimization, and learning models. They do not always fail in predictable ways, and vulnerabilities are not always solved with a simple patch.

Harriet shares her fascinating journey from studying physics and anthropology to working in data science, national security, and artificial intelligence, eventually discovering the world of adversarial machine learning — where attackers attempt to manipulate and disrupt AI systems themselves.

This conversation goes beyond the hype and explores what defenders, developers, and organizations need to understand as AI becomes a critical part of modern technology.

🤖 Why AI Security Matters More Than Ever

AI is becoming part of software development, business operations, healthcare, finance, critical infrastructure, and cybersecurity itself.

As adoption accelerates, organizations must move beyond simply asking:

“How can we use AI?”

and start asking:

“How do we secure AI?”

🧠 Understanding How AI Really Works

Harriet explains why machine learning systems are fundamentally different from traditional software.

AI systems are:

• Probabilistic rather than deterministic
• Dependent on training data quality
• Designed around optimization
• Continuously influenced by changing environments

Understanding these foundations is essential for anyone responsible for protecting AI.

🔓 The World of Adversarial Machine Learning

What happens when attackers stop targeting only applications and infrastructure…

…and start targeting the AI model itself?

The episode explores:

• Model manipulation
• Data poisoning
• AI weaknesses
• Training challenges
• Unexpected behaviors
• The difficulty of understanding model decisions

🛠️ How Do You Patch AI?

One of the biggest questions facing cybersecurity professionals today:

If AI learns something wrong, how do we fix it?

Traditional security follows a familiar process:

Find vulnerability → Apply patch → Reduce risk

AI changes that.

Sometimes protecting AI is not about fixing code.

It is about understanding and correcting behavior.

⚔️ AI for Security vs Security for AI

For years, organizations have focused on using AI to improve cybersecurity.

But now the challenge has expanded.

Cybersecurity needs AI.

But AI also needs cybersecurity.

As AI becomes part of everyday systems, security teams must understand how to protect the models, data, and decisions that organizations rely on.

🌍 Why AI Security Requires Different Skills

The future of AI security requires collaboration between:

• Cybersecurity professionals
• AI engineers
• Data scientists
• Researchers
• Risk leaders
• Policy experts

Building trustworthy AI means bringing these worlds together.

Security must be part of AI from the beginning.

🔹 Harriet’s journey from physics and anthropology into AI security

🔹 Working in data science and national security environments

🔹 Discovering adversarial machine learning

🔹 Founding Mileva Security Labs

🔹 Writing Practical AI Security with No Starch Press

🔹 Why AI vulnerabilities are different from software vulnerabilities

🔹 The importance of data quality and model training

🔹 Understanding probability and machine learning foundations

🔹 How attackers target AI systems

🔹 Why securing AI requires a new mindset

🔹 The future of AI safety and cybersecurity

🔹 Staying updated in a fast-moving industry

🔹 Building responsible and secure AI systems

💬 “Before we can secure AI, we first need to understand how it works.”

💬 “AI security is not always about fixing a bug. Sometimes it is about correcting a behavior.”

💬 “Cybersecurity needs AI, but AI also needs cybersecurity.”

💬 “The future is not just about building smarter AI — it is about building safer AI.”

00:00 – Introduction to Security by Default

01:03 – Harriet Farlow’s origin story

04:28 – From data science to cybersecurity

08:48 – Creating Mileva Security Labs

10:51 – Conferences, community, and writing Practical AI Security

17:28 – How AI has evolved

19:43 – Understanding machine learning models

21:43 – The challenge of patching AI systems

23:37 – Training data, quality, and user impact

25:23 – Why AI models can be difficult to understand

27:36 – AI and cybersecurity coming together

30:18 – Why AI fundamentals matter

32:04 – Practical examples and real-world AI security

33:38 – Staying updated in AI security

36:27 – Learning from the AI security community

38:08 – Ethics and responsible AI development

Harriet Farlow

Founder — Malevra Security Labs

Author — Practical AI Security

🔗 LinkedIn:

https://www.linkedin.com/in/harriet-farlow-654963b7/

📘 Practical AI Security — No Starch Press

https://nostarch.com

🎓 AI Fundamentals Course

https://harriethacks.com/course/

🎧 Security by Default Podcast

Exploring the people, stories, and ideas helping make technology safer.

Because security should not be an afterthought.

Security should be by default.

#SecurityByDefault #AISecurity #Cybersecurity #ArtificialIntelligence #MachineLearning #AdversarialML #AI #ResponsibleAI #SecurityResearch

Takeaways:

• The podcast episode discusses the importance of understanding AI security in the context of national security and its implications.
• Harriet's journey from a background in physics and anthropology to her current role in AI security demonstrates the interdisciplinary nature of the field.
• The conversation highlights the necessity for collaboration between AI developers and cybersecurity professionals to ensure secure AI systems.
• Listeners are encouraged to engage with various resources to stay informed about the rapidly evolving landscape of AI and cybersecurity.
• The significance of addressing the ethical considerations in AI development is emphasized throughout the discussion, focusing on empowering rather than replacing human effort.
• The episode underscores the idea that AI security is not merely about using AI for cybersecurity but also about securing AI systems from external threats.