Five AI safety sessions every founder should have on their TechCrunch Disrupt 2026 agenda
TechCrunch TechCrunch Events
TechCrunch Disrupt 2026 is lining up five AI safety sessions for founders. They focus on agents, enterprise security, robots, and why trust is the real product.
Based on reporting by TechCrunch, TechCrunch Events — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
TechCrunch Disrupt 2026 is making a simple point: the AI buzz has moved past demos, and the hard part now is trust. If founders want to put agents into company systems, ship enterprise software, or send robots into the real world, they need to think about safety and security as core product features, not add-ons.
That theme runs through five sessions spread across the AI Stage and the Real World AI Stage. One of the sharper ones looks at what happens after an enterprise actually starts using Claude. Anthropic’s Cat de Jong works directly with companies trying to put the model into critical workflows, which means she sees both the deals that turn into real value and the pilots that never seem to end. That’s a useful lens for anyone trying to sell AI into a business that has already heard enough about “transformation.”
Another session gets into agent security, where the obvious questions have messy answers: what should an agent be allowed to touch, what should it be able to do, and what happens when app-level permissions fall short? Okta’s Ric Smith and NanoCo’s Gavriel Cohen will take that problem down to the infrastructure level, which is where it belongs if software is actually going to take actions on its own.
The enterprise angle gets even tighter in a session on cloud security, governance, and observability. AWS security VP Rudy Mitra, Luta Security CEO Katie Moussouris, and cybersecurity veteran Wendy Nather will talk through what enterprises need when AI starts moving into more autonomous roles. And on the physical side, Shield AI’s Nathan Michael, General Motors’ Mikell Taylor, and Waabi’s Raquel Urtasun will tackle the question founders in autonomy keep running into: how do you know a system is safe enough when mistakes can affect vehicles, aircraft, industrial systems, or critical missions?
The final session goes straight at robotics, where the data problem is still brutal. NVIDIA’s Les Karpas will dig into how data pipelines, simulation, and foundation models might help robots get past a major scaling bottleneck. It’s a practical lineup, which fits the moment. The easy part is building something impressive. The hard part is convincing anyone to trust it enough to use it.
My take — AI-written commentary, not fact-checked reporting
This is the right obsession. AI founders love talking about capability, but the market keeps buying trust, and the bill always shows up in security, governance, and testing. The industry’s favorite trick is to call that “enterprise readiness,” which is just a nicer way of saying the product wasn’t finished yet.
Read more about this at: TechCrunch