Episode 53 · Season 1
AI Dev Tool Security: 4 Risks to Act On, 4 to Ignore
The AI security conversation has two volume settings: hand over the keys, or the sky is falling. On this episode of Cirrius Talk: Agentic AI, Tim Harting and Gavin Franklin give you the calibrated middle, a practical framework for AI dev tool security that holds up whether you are technical or not.
What you’ll learn:
- Which fears are overstated (internet access, file deletion, vendors training on your whole machine)
- The moderate risks worth monitoring: insecure code at scale, context and training data, malicious public repos, cost overruns
- The act-now threats: prompt injection, automated output without review, API keys in local files
- Six mitigation principles, from human in the loop to credential hygiene to spending limits
Who it’s for: Engineering leaders, architects, security owners, and business leaders deciding how to adopt AI dev tools without getting burned.
Guest + credibility: Tim Harting and Gavin Franklin work hands-on with agentic dev tools at Cirrius Solutions, advising teams on how to deploy them securely. They translate real implementation experience into rules you can apply immediately.
Primary CTA: Learn more at https://cirriussolutions.com/podcast/
Follow the show for more Cirrius Talk: Agentic AI episodes
Recommended Links & Resources
Tim Harting on LinkedIn
Gavin Franklin on LinkedIn
Greg Banks on LinkedIn
Jason Fowler Music
Cirrius Solutions
Agentic AI Playlist
Any Questions: Reach out to us at CirriusTalk@cirriussolutions.com
We cover
- Understanding AI Security Risks
- Overstated Fears in AI Security
- Moderate Risks of AI Implementation
- High-Risk Areas in AI Usage
- Mitigating AI Security Risks
- Practical Examples of Prompt Injection
Chapters
- 0:00Introduction to AI Conversations
- 2:27Understanding AI Security Risks
- 5:14Overstated Fears in AI Security
- 8:48Moderate Risks of AI Implementation
- 14:57High-Risk Areas in AI Usage
- 21:28Mitigating AI Security Risks
- 26:28Practical Examples of Prompt Injection
- 29:22Conclusion and Future Considerations
Resources and downloads
The artifacts Greg’s team built for this episode, free to download.
AI Dev Tool Security: 4 Risks to Act On, 4 to Ignore
A single-page reference that cuts through the hype around AI coding tools (Claude Code, Codex, Copilot), framing security risks with a traffic-light model: overstated fears to stop worrying about, moderate risks worth monitoring, and high-risk issues that demand action now. Pairs the risk breakdown with practical mitigation playbooks and a vendor-by-vendor look at security tracking and adoption policy.
Related episodes
Agentic AIClaude Code vs Wireframes: Building Interactive POCs in Hours, Not Weeks
Agentic AIAI Agent Strategy: 7 Deliverables to Lock In Before Development
Agentic AIAgentic AI – Evolution of Agent Hosting
The Military MinuteDean Robison – What If Veteran Hiring Isn’t Charity, but StrategyDean Robison
Salesforce at ScaleSalesforce Field Service Meets Agentforce: From Triage to Wrap UpTalk to us about AI agents
Keep going: See how we build AI agents for real businesses, Talk to us about a custom AI agent, More from the Cirrius blog