Cloud Security Podcast

Cloud Security Podcast is an interview show about how attacks land and where defenses fail, and its range goes well beyond cloud configuration despite the name. Browsers, endpoints, identity, the security operations center and AI systems all get attention. The conversations are specific. In one, a speaker describes adversaries using the same software as the enterprise so they can pass as an employee. Defenders are left knowing only that Zoom was running, not why remote control was handed over or what happened next. The same conversation argues that many alerts exist only because of dumb control points, and it asks whether a team can name its risky users and say why they are risky. Its most striking example has someone activating Meta AI inside WhatsApp and trying to paste patient records into it for a summary. The speakers call it unsanctioned AI in a sanctioned tool, and a HIPAA violation if it had gone through. The browser conversation is blunt. Patch management programs exist, one speaker says, but they do not work in reality, and an identity provider is not a firewall. Attackers phish through legitimate SaaS applications, and ClickFix tricks users into opening the Windows command prompt and pasting in a command. The attacks on JLR and Marks and Spencer come up, as do a campaign against Salesforce and a coalition posting on Telegram as Scattered Lapsus Hunters. AI agents are treated as a new attack surface. Every tool an agent runs is described as a chance for proprietary data to leak or for a malicious tool to send commands back, and the agent is compared to Ron Burgundy reading whatever is on the teleprompter. On adoption, organizations are said to fall into two camps: flat refusal, or floodgates already open with no idea what is being used. The proposed fix puts identity at the core, so that a device-bound, attributable identity replaces a probabilistic guess at the blast radius. The conversation behind 'Why AI Infrastructure is Harder to Secure Than Cloud' raises fourth- and fifth-party exposure through services like Bedrock and warns never to connect an MCP server directly to a database. Titles such as 'AI Agents for SOC: Hype Curve vs. Measurable ROI' and 'Is AI Hallucinations a Myth and the Real Threat from AI' show how often the SOC question comes back. One speaker describes multi-step attack campaigns finishing within minutes, too fast for a SOC to answer without AI. The AI SOC product Dropzone comes up, including its work with the Cloud Security Alliance on a report about AI-augmented SOCs. The forecast is restrained: fewer tier one analysts, more security architects, and no fully automated SOC. The closings point to a sister show, AI Security Podcast, and a newsletter at cloudsecuritynewsletter.com, and the host often signs off with one word: Peace.