AI is Rewriting Cybersecurity - Guardrails, regulation, and the point of no return w/ Joseph Carson
The Secure DisclosureShow Notes
Social engineering and phishing are evolving fast, and AI is making attacks harder to spot and quicker to scale. Joseph Carson joins the show to break down the biggest risks for defenders, from deepfakes and perfect-language phishing to rapid data analysis and malware that adapts in real time. The conversation also explores guardrails, regulation, and what AI can and cannot do well, plus a quick round of security themed “Would you rather” questions.Links: Linkedin: https://www.linkedin.com/in/josephcarson/Sponsored Link: https://www.aikido.dev/Chapters00:00 Intro: AI makes phishing harder to detect00:00:28 Welcome and Joe’s background00:01:29 Biggest risks: deepfakes and phishing at scale00:03:03 AI speeds up analysis of stolen data00:04:25 Lower barrier to entry and faster attacker learning00:05:31 Malware and campaigns adapting in real time00:06:28 Why “bad grammar” is no longer a phishing tell00:08:16 Can AI be creative, or is it just probability00:12:56 Guardrails, regulation, and the EU vs US vs China approaches00:29:30 Would you rather: security tradeoffs and tool choices#podcast #thesecuredisclosure #cybersecurity
More from The Secure Disclosure
AI coding agents ship code at record speed, and ship vulnerabilities just as fast. Edouard Viot, Co-founder and CTO at Symbiotic Security, explains why scanning in CI is already too late, how AI harnesses and system prompts build real-time guardrails, whether deterministic SAST is dead, and why malicious skills and plugins are the next big attack surface.Edouard Viot is the Co-founder and CTO at Symbiotic Security. Symbiotic Security builds end-to-end security for AI-assisted software development, securing code as it is generated inside coding agents rather than after the fact. Learn more at https://symbioticsec.aiThis episode is sponsored by Aikido Security, the no-nonsense software security platform. Start for free at https://aikido.dev
Phishing remains cybercrime's primary entry point, but outdated templates and punitive culture fail to stop AI-driven social engineering. In this episode of The Secure Disclosure, host Mackenzie Jackson sits down with Simon Bauwens, co-founder and CEO of Outkept. Simon explains why spotting phishing is a perishable reflex rather than a learned skill, how MFA fatigue leaves organizations vulnerable, and why Outkeep pays a crowdsourced community of ethical phishers based on simulation click rates. Finally, they explore building positive security cultures and play our signature Would You Rather challenge.
We are handing non-deterministic AI agents static API keys, then acting surprised when production databases get wiped.
As autonomous AI agents execute database queries and trigger APIs at machine speed, static permission models are collapsing. In this episode of The Secure Disclosure, host Mackenzie Jackson sits down with David Mytton, CEO of Arcjet, to explore the shift from traditional network firewalls to application-native runtime security.David shares insights on managing non-deterministic non-human identities, mitigating prompt injection risks, and why granular authorization is essential for agentic workflows. He also debunks common myths surrounding data center water and energy consumption from his sustainable computing research, before facing a chaotic round of Would You Rather.
Is AI wiping out entry-level cybersecurity jobs? On this episode of The Secure Disclosure, host Mackenzie Jackson sits down with veteran and advisory CISO Michael Farnum. They dive into why teams must presume exploitability, how newcomers can break into security using home labs instead of just college degrees, and why conference accessibility matters.