The Next Chapter of AI Is Inside Our Software
a16z · 2026-09-26 · 55 min
https://www.youtube.com/watch?v=TLJNJDf2XGoVideo summary
Aaron Levie and Martin Casado debate AI regulation, agent cybersecurity, and a software shift toward probabilistic models.
Box CEO Aaron Levie, a16z’s Martin Casado, and Steven Sinofsky challenge calls to pace or regulate AI before its risks are clearly defined. They argue that extinction claims and vague safety rhetoric can invite heavy-handed controls, while specific, demonstrated risks—especially cybersecurity—offer a more productive basis for action. Drawing on the internet’s early virus outbreaks, aviation safety standards, and computer-crime laws, they say industries often learn how to manage new technologies through real-world experience. The conversation turns to AI agents: swarms could act like thousands of roaming users, making stronger authentication, API monitoring, and more granular permissions essential. The panel also warns that Europe could export prompt-heavy AI rules, and predicts AI will be a major issue in the 2028 election. In the final segment, they discuss models that classify inputs and return options rather than generating long answers, making AI faster and cheaper to integrate into conventional software—and shifting innovation beyond the model labs.
Chapters
- 0:00Pacing the Frontier: Levie Backs AI Sandboxing but Warns Against Regulatory Capture
- 3:50Species Extinction Talk: Casado Challenges the 10% Claim and Calls for a Clear Lab Position
- 6:49The Nationalization Question: Extinction Claims, Lab HR Concerns, and Government Regulation
- 12:082028 as the AI Election: Pro-AI Advocates Struggle to Counter Extinction-Risk Rhetoric
- 15:562028 as the AI Election: Nonzero Risk, BSL-4 Controls, and the Regulatory Fate of AI
- 20:112028 as the AI Election: FINRA, Congress’s Internet Regrets, and Lessons from Early PC Worms
- 24:232028 as the AI Election: The 1983 GTE Intrusion and 1986 Computer Crime Law
- 28:39Why Cybersecurity Has Historically Rejected Practical Solutions: Heat Exfiltration and TEMPEST Leaks
- 32:30The Craziest Covert Channel Ever Seen: Reconstructing CRT Screens Through Window Light
- 36:26The Craziest Covert Channel Ever Seen: Secure-by-Design Systems and Granular Agent Permissions
- 41:05The Craziest Covert Channel Ever Seen: GDPR-Style AI Prompts and Warning Fatigue
- 44:31The Craziest Covert Channel Ever Seen: From a Compromised Stanford Oscilloscope to AI Regulation
- 48:38LLMs as Decision Engines vs Chatbots: Jev Returns Percentages for Probabilistic Software
- 53:01Why the Labs Haven't Built This: Being-Oriented Labs and Innovation Beyond the Model
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
How Cursor Built One of AI’s Fastest-Growing Companiesa16zCursor bet on the human-model interface, grew rapidly without early sales hires, and later expanded through Graphite and founder acquisitions.
Why Top Founders Are Racing Into AI Infrastructurea16za16z’s Machine Age Fund targets AI infrastructure as GPU supply is booked to 2028 and memory demand needs three years of capacity.
Why AI Demand Is Outrunning Compute Supplya16zGavin Baker argues AI compute demand will outpace supply, with sub-one-year infrastructure paybacks and orbital data centers approaching.
What Today’s Best Models Still Can’t Do in Matha16zDaniel Litt praises AI’s Erdős unit-distance result but warns that proofs alone cannot replace mathematical understanding or human curiosity.
Inside Moderna’s Biggest mRNA Test Since COVIDa16zModerna and Merck’s personalized mRNA vaccine beat Keytruda alone in a Phase 3 melanoma trial, after 1,000 cancer-vaccine trials failed.
Related analyses
To Regulate AI Effectively, Focus on How It’s Useda16z Deep DivesMartin Casado argues AI laws should target harmful uses, while regulatory uncertainty pushes US startups toward Chinese open-source models.
Cybersecurity in the Agentic Era | Deep Dives with a16za16z Deep Divesa16z, Neo, and Cotool warn that AI agents are breaking traditional cybersecurity defenses as 50% of enterprise apps go agentic.
Braintrust CEO on Where Engineering Actually Matters in AIa16z Deep DivesBraintrust CEO Ankur Goyal says AI engineering should focus on evals and harnesses, while SQL beat Bash in his agent benchmark.
How The Internet’s Favourite AI Employee Went RogueColdFusionOpenClaw promised a capable AI assistant but exposed private data, compromised 4,000 developer machines, and helped drive a $1 billion mortgage fraud investigation.
Why AI Agents Need Context | Deep Dives with a16za16z Deep DivesFivetran CEO George Fraser says AI agents need centralized business data, while warning AI-native rivals could outpace SaaS incumbents.