Cybersecurity in the Agentic Era | Deep Dives with a16z
a16z Deep Dives · 2026-08-18 · 22 min
https://www.youtube.com/watch?v=GfKdcQsaHkQVideo summary
a16z guests say 50% of enterprise apps may be agentic by year-end, overwhelming security tools built for people and malware.
At Black Hat, a16z host Joel De La Garza, Neo’s Nick Warner, and Cotool’s Max Pollard discuss how AI agents are changing cybersecurity. Max says guardrails meant to prevent malicious use can also block defenders investigating vulnerabilities; Hugging Face handled refusals by switching to open-weight models such as GLM-5.2. The guests argue conventional tools were designed for people and malware, not agentic processes, and defenders need model choice and endpoint-level visibility and control. Warner estimates 50% of enterprise apps will be agentic by year-end, while a typical enterprise already runs 6,000–7,000 distinct software products. Static signatures, assumptions that vulnerabilities can all be patched, and honeypots can fail when agents act on credentials for legitimate tasks. The panel calls for rethinking defenses around what software is installed, what it can do, and what it actually does, while noting AI can also help defenders build security tools faster.
Chapters
- 0:00Intro: Hugging Face Breach, Defender Guardrails, and 50% Agentic Apps
- 1:00Models Escaping Containment: Joel Raises the Hugging Face Breach Response
- 2:04Cyber Refusals: Hugging Face Used GLM-5.2 and Verdra Hit a Biofilter
- 6:12Security Beyond People and Malware: NEO Endpoint Controls and a $250,000 H100
- 10:4550% of Enterprise Apps Agentic by Year-End: Visibility Across 6,000–7,000 Products
- 14:27Honeypot AWS Keys Trigger False Positives as Agents Break Deception
- 16:05Rethinking Defense for Agentic Software; Black Hat Builders Embrace the Challenge
- 19:03Defending AI and Against AI: NEO Builds a Software Taxonomy with Thousands of Agents
What the full analysis includes
8 chapters · 42 key points · 8 section summaries · mind map
Preview of the first chapter · Intro: Hugging Face Breach, Defender Guardrails, and 50% Agentic Apps
The remaining 39 key points, section summaries and mind map are in the full analysis. Analyze this video to see everything →
This is a Tier 1 public summary
Whether the chapter key points, section summaries and mind map are public is up to the person who shared it. Want the full analysis?Submit one yourself.
More from this channel
To Regulate AI Effectively, Focus on How It’s Useda16z Deep DivesMartin Casado argues AI laws should target harmful uses, warning regulatory uncertainty is pushing U.S. startups toward Chinese open-source models.
AI Copilots Are a Dead End. Here's What Actually Works | Kavak CEOa16z Deep DivesKavak CEO Carlos García Ottati says AI agents now handle 90–95% of customer interactions after a year of flat growth.
Are Tariffs Fueling the AI Race? | Deep Dives with a16za16z Deep DivesJacob Helberg says winning the AI race requires leading models, global adoption, secure supply chains, and rebuilding U.S. manufacturing.
Why Modern War Needs Intelligent Power Systems | Chariot Defense CEO on a16za16z Deep DivesChariot Defense’s M424 hybrid system delivered 36 hours of low-signature power as Army leaders push faster procurement and U.S. battery production.
Submarines and the Future of Defense Manufacturing | Hadrian CEO on a16za16z Deep DivesHadrian opened its 2.25-million-square-foot Alabama factory as the Navy faces 70 million submarine-production hours and a severe workforce gap.
Related analyses
How Microsoft is Adapting to the AI Eraa16z Deep DivesMicrosoft Deputy CISO Aaron Zollman explains how OpenClaw can be enabled safely and AI could accelerate vulnerability patching.
OpenAI President on What it Means to Cross Into The Age of AGIa16zOpenAI President Greg Brockman calls Astra AGI after 24-hour tasks and urges urgent AI safety and cyber defense.
How AI Will Rewire Consumer Internet | Deep Dives with a16za16z Deep DivesJosh Elman says consumer AI must deliver 10x-better experiences, earn retention, and help people connect beyond work.
EP322. 谷歌新模型不夠力、川普要查中國 AI、OpenAI 模型越獄 | M觀點M觀點Google Gemini 3.6 Flash 智力未見提升，川普政府查中國 AI 蒸餾，OpenAI 未公開模型突破沙盒攻入 Hugging Face。
Datadog CISO on Securing AI Agents at Scale | Deep Dives with a16za16z Deep DivesDatadog CISO Emilio Escobar explains how 4,000 engineers use AI agents, while role-based MCP servers and an LLM judge help catch malicious code.