Absolute AppSec

In episode 315 of Absolute AppSec, Ken Johnson and Seth Law discuss the rapidly evolving challenges of securing software in an era of AI-assisted development. The hosts provide updates on their "Harnessing LLMs for Application Security" training, noting that the field is changing so fast that they must constantly update their exercises to include new agents and advanced tools like Claude Code. A primary concern raised is the "naivete" of many new security tools, where prompts are often automatically generated by AI rather than expertly crafted, causing a loss of essential nuance. The hosts also warn against AI companies building security products without specialized expertise, citing a zero-click exploit in the "Comet" AI browser that could exfiltrate sensitive secrets via calendar summaries. As development teams now ship code at "AI speed," the hosts argue that traditional AppSec methods are too slow, necessitating a strategic pivot toward automated design reviews, governance, and observability rather than just chasing individual vulnerabilities. Despite the inherent risks and the ongoing difficulty of managing AI reasoning drift, they remain optimistic that these tools can eventually unlock more efficient, hands-off AppSec workflows if managed with proper guardrails and deterministic oversight.

In this episode, the hosts discuss the seismic shift in the application security landscape triggered by the rise of Large Language Models (LLMs) and Anthropic’s "Claude Code". They highlight the massive economic repercussions of these AI advancements, noting that billions in market value were wiped from traditional cybersecurity stocks as investors begin to believe frontier models might eventually write perfectly secure code. The hosts critique the industry's historical reliance on "checkbox" compliance tools like SAST, DAST, and SCA, arguing that these "archaic" methods are being replaced by AI-native strategies capable of reasoning through complex logic flaws. While they acknowledge that AI can suffer from "reasoning drift" and still requires deterministic validation to avoid false positives, they emphasize that security professionals must adapt by building custom "skills" and focusing on governance and observability. The discussion concludes that as developers move to "AI speed," the traditional role of the AppSec professional is evolving into a "Jarvis-like" orchestrator who manages automated workflows and infuses institutional knowledge into AI agents to maintain oversight without slowing down production.

Ken Johnson and Seth Law examine the intensifying pressure on security practitioners as AI-driven development causes an unprecedented acceleration in industry velocity. A primary theme is the emergence of "shadow AI," where developers utilize unauthorized AI coding assistants and personal agents, introducing significant data classification risks and supply chain vulnerabilities. The discussion dives into technical concepts like AI agent "skills"—markdown files providing specialized directions—and the corresponding security risks found in new skill registries, such as malicious tools designed to exfiltrate credentials and crypto assets. The hosts also review 1Password’s SCAM (Security Comprehension Awareness Measure), highlighting broad performance gaps in an AI's ability to detect phishing, with some models failing up to 65% of the time. To manage these unpredictable systems, the hosts advocate for a shift toward high-level validation roles, emphasizing the need for Subject Matter Expertise to combat "reasoning drift" and maintain safety through test-driven development and periodic "checkpoints". Ultimately, they conclude that while AI can simulate expertise, human oversight remains vital to secure the probabilistic nature of modern agentic workflows.

In episode 312 of Absolute AppSec, the hosts discuss the double-edged sword of "vibe coding", noting that while AI agents often write better functional tests than humans, they frequently struggle with nuanced authorization patterns and inherit "upkeep costs" as foundational models change behavior over time. A central theme of the episode is that the greatest security risk to an organization is not AI itself, but an exhausted security team. The hosts explore how burnout often manifests as "silent withdrawal" and emphasize that managers must proactively draw out these issues within organizations that often treat security as a mere cost center. Additionally, they review new defensive strategies, such as TrapSec, a framework for deploying canary API endpoints to detect malicious scanning. They also highlight the value of security scorecarding—pioneered by companies like Netflix and GitHub—as a maturity activity that provides a holistic, blame-free view of application health by aggregating multiple metrics. The episode concludes with a reminder that technical tools like Semgrep remain essential for efficiency, even as practitioners increasingly leverage the probabilistic creativity of LLMs.

Ken Johnson and Seth Law examine the profound transformation of the security industry as AI tooling moves from simple generative models to sophisticated agentic architectures. A primary theme is the dramatic surge in development velocity, with some organizations seeing pull request volumes increase by over 800% as developers allow AI agents to operate nearly hands-off. This shift is redefining the role of Application Security practitioners, moving experts from manual tasks like manipulating Burp Suite requests to a validation-centric role where they spot-check complex findings generated by AI in minutes. The hosts characterize older security tools as "primitive" compared to modern AI analysis, which can now identify human-level flaws like complex authorization bypasses. A major technical highlight is the introduction of agent "skills"—markdown files containing instructions that empower coding assistants—and the associated emergence of new supply chain risks. They specifically reference research on malicious skills designed to exfiltrate crypto wallets and SSH credentials, warning that registries for these skills lack adequate security responses. To manage the inherent "reasoning drift" of AI, the hosts argue that test-driven development has become a critical safety requirement. Ultimately, they warn that the industry has already shifted fundamentally, and security professionals must lean into these new technologies immediately to avoid becoming obsolete in a day-to-day evolving landscape.

In this episode of Absolute AppSec, hosts Ken Johnson and Seth Law interview Mohan Kumar and Naveen K Mahavisnu, the practitioner-founders of Aira Security, to explore the critical challenges of securing autonomous AI agents in 2026. The conversation centers on the industry's shift toward "agentic workflows," where AI is delegated complex tasks that require monitoring not just for access control, but for the underlying "intent" of the agent's actions. The founders explain that agents can experience "reasoning drift," taking dangerous or unintended shortcuts to complete missions, which necessitates advanced guardrails like "trajectory analysis" and human-in-the-loop interventions to ensure safety and data integrity. A significant portion of the episode is dedicated to the security of the Model Context Protocol (MCP), highlighting how these integration servers can be vulnerable to "shadowing attacks" and indirect prompt injections—exemplified by a real-world case where private code was exfiltrated via a public GitHub pull request. To address these gaps, the guests introduce their open-source tool, MCP Checkpoint, which allows developers to baseline their agentic configurations and detect malicious changes in third-party tooling. Throughout the discussion, the group emphasizes that as AI moves into production, security must evolve into a proactive enablement layer that understands the probabilistic and unpredictable nature of LLM reasoning.

In this episode of Absolute AppSec, Nathan Hunstad, Director of Security at Vanta, discusses the intersection of security policy, governance, and technical defense. Drawing on his unique background in political science and the Minnesota state legislature, Hunstad argues that policy acts as the essential "conductor" for an organization's security tools. A major theme of the conversation is the challenge of compliance for startups, with the group advising founders to prioritize business survival and basic security hygiene—like password managers and IAM—before pursuing intensive certifications like SOC 2. The discussion also explores how AI is accelerating both development velocity and the ability to automate tedious security questionnaires. Furthermore, Hunstad contrasts the security posture of modern, cloud-native startups against legacy enterprises, noting that older organizations often struggle with "dark corners" of un-inventoried, vulnerable legacy tech. The episode concludes with a critique of outdated authentication standards, specifically advocating for the removal of mandatory password rotation in favor of NIST-aligned, phishing-resistant MFA.

Ken Johnson (cktricky on social media) and Seth Law are happy to announce a special episode of Absolute AppSec with Avi Douglen (sec_tigger on X), long-time OWASP Global Board of Directors member, founder and CEO of Bounce Security and co-author of the Threat Modeling Manifesto. The conversation ranges from Application Privacy related to Application Security, to participating in meetups and conferences, and finally OWASP as an Avi's experience as a board member.

In episode 307 of Absolute AppSec, hosts Ken and Seth conduct a retrospective on the application security landscape of 2025. They conclude that their previous predictions were largely accurate, particularly regarding the rise of prompt injection, AI-backed attacks, and the industry-wide shift toward per-token billing models. A major theme of the year was the solidification of supply chain security as a critical pillar of AppSec, driven by notable incidents such as Shai Hulud and React for Shell. The hosts also share insights from their four-day training course on utilizing LLMs for secure code review, noting that while AI development is becoming more prevalent, most practitioners are still in the nascent stages of building custom tooling. Much of the discussion focuses on the Model Context Protocol (MCP); while it offers significant value for agentic workflows, the hosts criticize its current lack of robust security controls, specifically highlighting issues with OAuth implementations and short timeouts in existing clients. Finally, they discuss how the industry is moving toward a more nuanced balance between deterministic tools like Semgrep and the probabilistic creativity of LLMs to increase efficiency in security consulting.

Given the spate of recent npm news stories, we've arranged a topical show with software supply-chain security researcher and npm hacker Paul McCarty (find Paul on bsky https://bsky.app/profile/6mile.githax.com) . Paul is currently a researcher with Safety (https://getsafety.com/) and has a background in security including work at John Deere, Boeing, Regence Blue Cross/Blue Shield, NASA Jet Propulsion Lab, the US Army, and the Queensland Government. He's also spent twenty some odd years helping startups with security practices, and is a maintainer of the Open Source Malware project. In addition, Paul has been long time friend of the show, contributing his insights to the Absolute AppSec community slack in addition to frequently writing up his research at the SourceCode RED blog: https://sourcecodered.com/blog.