Aaron Levie, Steven Sinofsky & Martin Casado: How Do You Secure a World of AI Agents?
September 26, 2026
AI Summary
5 min readHow Do You Secure a World of AI Agents?
The AI safety debate has a fundamental problem: the labs and the security community keep talking past each other. When Anthropic CEO Dario Amodei published a post about "pacing" AI development—slowing releases to prioritize safety—the substance was reasonable, but the atmospherics were not. As one participant put it, "An employee is like 10% chance of species extinction. The post is very reasonable, but the atmospherics are not." This tension between existential risk rhetoric and practical engineering concerns runs through the entire conversation about how to secure a world where AI agents operate across the software we use every day.
The Pacing Problem
The concept of "pacing" AI development has become a flashpoint. On its face, the idea that frontier labs should build in the safest way possible—with strong governance, security, and alignment—is uncontroversial. No one disagrees that AI diffusion requires trustworthy products. But the term itself creates confusion. As one participant noted, pacing is "orthogonal to security. You can very slowly build a nuclear weapon. That doesn't make anybody feel better."
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of a16z Show
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **The Core Debate: Pacing, X-Risk, and Atmospherics** - The panel opens by dissecting the recent "pacing" proposal from frontier labs, arguing that the substance (security best practices) is reasonable, but the messaging and surrounding "atmospherics" (e.g., employee statements about 10% extinction risk) are broken.
- 2 (08:06) **Why "Pacing" is a Dangerous Word** - The conversation shifts to why the "pacing" framing is politically naive and strategically risky, as it invites government intervention without a clear goal.
- 3 (13:27) **The X-Risk Trap: Why Labs Must Take a Position** - The discussion focuses on the strategic necessity for frontier labs to directly address the existential risk question rather than leaving it ambiguous.
- 4 (17:42) **Historical Lessons: How Tech Industries Get Regulated** - The panel draws on history to predict AI's regulatory future, from the movie ratings board to the financial industry.
- 5 (23:44) **The Security Blind Spot: Learning from the Internet's Chaos** - The conversation pivots to concrete, known risks, arguing that the current discourse should focus on cybersecurity, not philosophical X-risk.
- 6 (30:51) **The Threat of AI Agents: Swarms, APIs, and Internal Security** - The discussion identifies a novel, concrete risk: how autonomous AI agents will fundamentally break existing internal security models.
- 7 (36:43) **The Solution: A Renaissance in Secure Systems Design** - The panel argues that AI itself may force and enable a complete rethinking of the software stack to be secure by design.
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
Erik Torenberg sits down with Box CEO Aaron Levie, and a16z’s Martin Casado, and Steven Sinofsky to debate how the AI industry should think about safety, security, and regulation as increasingly capable agents move into the real world.
They argue that much of today’s conversation is happening before we have clearly defined the risks we’re trying to regulate. Drawing on earlier waves of computing, from computer viruses and the early internet to aviation and automobiles, they ask what AI can learn from industries that developed safety standards only after understanding how their technologies actually failed.
The conversation then gets concrete: agents don’t get tired, can operate at enormous scale, and can probe systems in ways human employees never could. That could require rethinking permissions, authentication, operating systems, and the security stack itself. They also discuss why AI innovation may increasingly move beyond the frontier labs and into the software built around the models.
Resources:
Follow Aaron Levie on X: https://x.com/levie
Follow Martin Casado on X: https://x.com/martin_casado
Follow Steven Sinofsky on X: https://x.com/stevesi
Stay Updated:
Find a16z on YouTube: YouTube
Find a16z on X
Find a16z on LinkedIn
Listen to the a16z Show on Spotify
Listen to the a16z Show on Apple Podcasts
Follow our host: https://twitter.com/eriktorenberg
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.