AI Employees Are Here. How to Secure AI agents. | Ankur Shah | Season 1 EP 8

Value of the Information & Strength of the Argument

The video provides valuable insights into the current state of AI agent security, drawing on the guest’s experience with over 500 enterprises. The argumentation is solid, with practical recommendations grounded in real-world incidents and research. The discussion on MCP supply chain attacks and the need for runtime controls is particularly compelling. The guest’s emphasis on enabling rather than blocking AI adoption is a nuanced perspective that adds depth to the conversation.

Scientific Rigor, Source Quality, Title Accuracy

The video demonstrates scientific rigor by referencing specific research and incidents, such as the Claude Code source leak and MCP supply chain attacks. The guest cites his company’s research and industry reports, which adds credibility. The title accurately reflects the content, focusing on securing AI agents. The discussion is well-structured and stays on topic, with no significant digressions.

The title accurately reflects the content, which focuses on securing AI agents in enterprise environments.

8/10

The video features a cybersecurity expert with extensive industry experience, discussing current threats and practical mitigation strategies. Claims are supported by references to published research and industry reports. However, some statements are anecdotal and lack detailed evidence.

• 00:00The AI security threat keeping CISOs awake at night
• 02:47Why enterprises are suddenly treating AI agents like employees
• 05:31The rise of agentic social engineering and Shadow AI
• 09:25The AI tools spreading faster than anyone expected
• 12:29The 3 steps every security leader should take today
• 15:16The MCP security problem most teams are missing
• 20:04The questions every CISO should be asking right now
• 24:16The fastest technology adoption in history
• 28:31One thing every security leader should do this week

• Straiker — Company website of the guest's company, providing context on their security solutions.
• Straiker STAR Labs Research — Research page with publications on AI-native attacks, including MCP supply chain attacks.
• Claude Code Source Leak — Blog post discussing the Claude Code source leak and its security implications.
• NomShub — Cursor remote tunnel breakout — Blog post about a sandbox breakout in Cursor, highlighting agent security risks.
• From Inbox to Wipeout — Perplexity Comet — Blog post about an AI browser wiping a Google Drive, illustrating agent risks.
• SmartLoader clones Oura Ring MCP — Blog post about a supply chain attack via a cloned MCP server.
• Why 94% of AI Agents Are Vulnerable to Prompt Injection — Blog post discussing the prevalence of prompt injection vulnerabilities in AI agents.
• OWASP's First Top 10 for Agentic AI — Blog post summarizing OWASP's Top 10 for agentic AI, a key reference for security.
• 10 Hard-Won Lessons Building an AI Security Company — Blog post with lessons learned from building an AI security company.
• Securing Agentic AI in Banking & Financial Services — Blog post on securing AI agents in the banking sector.
• Straiker named to CB Insights 2026 AI 100 — Announcement of Straiker's recognition, adding credibility to the guest.

• OWASP Top 10 for LLM Applications — Aligns with the video's emphasis on prompt injection and agent security.
• NIST AI Risk Management Framework — Provides a broader framework for managing AI risks, complementing the video's recommendations.

• www.evabenn.com