Summary
In this a16z podcast episode, a group of investors and operators debate how AI should be governed and where the next breakthroughs will come from. The conversation starts with the recent discussion of “pacing the frontier” and a widely read post from Anthropic CEO Dario Amodei on security and safety practices at leading labs.
The panel largely agrees that frontier labs should build with strong governance, sandboxing and testing, since enterprises will not adopt AI they cannot trust. The disagreement is over how policymakers might use those arguments. Speakers worry about regulatory capture, premature rules that fail to address real risk, and political backlash that could lead to data center bans, and they argue that the public mood around AI safety has become distorted.
The hosts also draw historical comparisons to the early internet, when worms, viruses and outages caused serious economic damage without stopping its growth, and to how Washington felt it had missed its chance to shape that era. Other themes include the growing rift between labs and the security community, the new authentication and API tracking that agent swarms require, European leadership in tech antitrust, and the idea that innovation may be shifting away from the large labs.
📺 Source: a16z · Published September 26, 2026
🏷️ Format: Podcast
