CyberWire Daily Windows servers under siege
7 snips
Oct 28, 2025 
David Moulton, the host of the Threat Vector segment, chats with cybersecurity experts Sarit Tager and Krithivasan Mecheri from Palo Alto Networks. They tackle the pressing issue of securing modern development in the age of AI, discussing challenges like shifting security practices and the importance of AI training for developers. The trio also dives into overlooked vulnerabilities in AI-generated code, highlighting real-world implications like the Doritos misidentification incident. A conversation not to be missed for anyone interested in the intersection of AI and cybersecurity!
AI Snips
Chapters
Transcript
Episode notes
AI Shifts Developer Responsibility
- AI coding shifts responsibility and knowledge away from individual developers to agents, creating surprise risks in generated code.
- New agent-related attack surfaces and unmanaged MCPs increase supply-chain and runtime exposure.
Scan Before You Generate
- Integrate scanning and security checks as prerequisites during code generation rather than after code is written.
- Ensure agents adhere to security requirements and block unsafe suggestions before they reach repos.
Agents Create New Supply-Chain Nodes
- Tooling ecosystems (LLMs, MCPs, agent platforms) become new supply-chain nodes that may be unmanaged and insecure.
- Unapproved external servers used by agents can execute actions outside an organization's approved whitelist.