The Critical Safety Challenge Facing AI Entrepreneurs

When Innovation Meets Responsibility

At first glance, the question might seem academic: would you trust an AI agent with access to your company's systems? What about putting employees in autonomous vehicles or deploying robots that must navigate unpredictable physical environments?

But as we've seen over the past few years, these aren't hypothetical questions for AI founders anymore. They're urgent business realities that define whether companies succeed or fail in an increasingly complex technological landscape.

The challenge isn't just about building better models or more capable agents – it's about building trust.

At TechCrunch Disrupt 2026, five sessions across the AI Stage and Real World AI Stage will confront these critical challenges head-on. These aren't theoretical discussions about abstract concepts; they're real-world insights from industry leaders who are actively wrestling with these issues in their own companies.

From securing autonomous agents to getting enterprise AI into production, we'll explore what it truly takes to build artificial intelligence that people and enterprises will actually trust and use. For founders looking to navigate the path from innovation to real-world adoption, this agenda is essential reading.

Anthropic's Enterprise Perspective

Some enterprises are already seeing measurable value from AI implementations. Others remain stuck in pilot mode 18 months later. Why does one organization succeed while another struggles?

Anthropic Head of Applied AI Cat de Jong brings a unique perspective to this question. She works directly with enterprises that are putting Claude into critical workflows and sees firsthand what separates successful deployments from those that stall.

In "What Anthropic Sees When Enterprises Actually Deploy Claude," de Jong will take the AI Stage to explore what happens when companies move from experimenting with AI to putting it into production. For founders trying to sell AI into the enterprise, her experience offers a firsthand look at why some deployments deliver real value while others remain stuck in pilot mode.

What makes the difference? Is it the technology itself, or is it something more fundamental about how organizations approach implementation and integration? The answer has implications for every founder building AI products that need to scale beyond initial experiments.

The Hidden Threat of Agent Security

An AI agent that can take action introduces an entirely new set of security problems. What should it have access to? What should it be allowed to do? And what happens when application-level permissions aren't sufficient?

This is the core challenge that Okta President of Products and Technology Ric Smith and NanoCo co-founder and CEO Gavriel Cohen will address in their session "The Agent Security Problem Nobody Is Talking About."

They'll examine agent security at the infrastructure level, including weaknesses in application-level permission models and architectural decisions that founders must consider when deploying agentic AI. This is particularly critical as we move toward more autonomous systems that operate with increasing independence.

As AI agents become more sophisticated and take on greater responsibilities, the stakes for security grow exponentially. The session will help founders understand what vulnerabilities they might be overlooking and how to build robust security frameworks from the ground up.

Enterprise Security Complexity

Getting an AI product through the enterprise door requires more than just a compelling demo. Security, governance, and observability all become part of the conversation when companies consider putting AI into critical systems.

In "Securing the AI Enterprise: Why the Cloud Just Got a Lot More Complicated," AWS VP of Security Services Rudy Mitra; Luta Security CEO Katie Moussouris; and cybersecurity veteran Wendy Nather will examine the infrastructure that enterprises need as AI takes on more autonomous roles.

This session is particularly relevant for founders building enterprise AI products. It provides insights into what happens when innovation meets the security requirements of organizations that have been protecting critical systems for decades.

The reality is that AI security isn't just about preventing unauthorized access – it's about ensuring accountability, traceability, and compliance with increasingly complex regulatory environments. For founders, this means thinking beyond traditional cybersecurity approaches to consider how AI systems can be designed with safety as a fundamental principle from the start.

Building Systems When Failure Is Not an Option

AI that operates in the physical world raises the stakes significantly. A mistake doesn't stay on a screen – it can affect vehicles, aircraft, industrial systems, and critical missions.

On the Real World AI Stage, Shield AI Chief Technology Officer Nathan Michael; General Motors Director of Robotics Strategy Mikell Taylor; and Waabi founder and CEO Raquel Urtasun will bring perspectives from defense and autonomous systems to one of the toughest questions hard tech founders face: How do you know when an autonomous system is safe enough to deploy?

The conversation in "Building AI Systems When Failure Is Not an Option" will dive into creating a safety culture, testing and validating AI systems, navigating regulatory hurdles, and building companies that can earn trust when the consequences of failure are physical.

This session represents the most demanding application area for AI – where human lives are literally on the line. It provides crucial insights into how we balance innovation with absolute safety requirements in critical applications.

Robots Waiting for Their ChatGPT Moment

Robots have a data problem. They don't have access to the massive pools of training data that helped accelerate advances in language models and self-driving vehicles.

This gap is one of the biggest obstacles to scaling physical AI. In "Robots Are Waiting for Their ChatGPT Moment. Here Is What Is Standing in the Way," Nvidia Inception Global Head of Physical AI Les Karpas will explore how data pipelines, simulation environments, and foundation models could help close this gap and what it will take to build, test, and deploy more capable and reliable robots.

For founders working in physical AI, this session tackles a big question: What will it take for robotics to reach its own ChatGPT moment and earn the trust needed to put those systems to work in the real world?

The key insight here is that robotics faces different challenges than other AI applications. While language models can learn from vast amounts of text, physical AI requires very different types of data – often limited by expensive testing, simulation, or real-world deployment.

Trust as the Ultimate Differentiator

What connects all five sessions is a common thread: trust. Whether it's trust in enterprise systems, autonomous agents, or physical robots, this fundamental element determines whether AI innovations become transformative technologies or remain confined to research labs.

As we move forward, the companies that will succeed are those that understand that AI safety isn't just an afterthought to add at the end of development. It's a core design principle that must be woven into every aspect of how AI systems are developed and deployed.

The sessions at Disrupt 2026 represent a critical juncture in the AI industry's evolution. They're not just about technical solutions – they're about building confidence in technology that will shape our world for decades to come. For founders, this is an opportunity to learn from leaders who are already addressing these challenges head-on.

Building a smarter model, a more capable agent, or a robot that does something no one has done before – all of these are impressive achievements. But getting customers to trust it enough to deploy it may be one of the hardest parts of bringing it to market. These sessions offer insights into how to bridge that critical gap.

At TechCrunch Disrupt 2026, we'll see firsthand how industry leaders are tackling these safety, security, and reliability challenges that can stand between breakthrough technology and real-world adoption. This is where the future of AI will be shaped – not just by what we can build, but by how responsibly we choose to deploy it.