Nvidia CEO Jensen Huang gives the keynote address at the company's annual GTC developers conference in San Jose, California, on March 16. [Photo/AFP]

The rise of OpenClaw, the open-source autonomous AI agent that has taken the tech world by storm, was a hot topic at Nvidia's annual conference this week, with industry leaders hailing it as a landmark shift in artificial intelligence while issuing clear warnings about the security risks it carries.

Held from Monday to Thursday in San Jose, California, Nvidia's GPU Technology Conference, or GTC, brought together some of the most influential voices in the global AI industry. Across panel discussions and keynote stages, industry leaders said OpenClaw is advancing AI capabilities significantly, from answering questions to taking action.

Launched in November 2025 by Austrian developer Peter Steinberger, OpenClaw is capable of handling daily tasks, such as clearing inboxes, sending emails, managing calendars and checking users in for flights. It has since become one of the fastest-growing projects in the history of GitHub, the world's most widely used developer platform. It's also gone viral globally, including in China, where users coined the phrase "raising a lobster" to describe training the AI assistant.

"There's been a lot of developer creativity coming out of that community. It's been something that's absolutely fascinating," Ali Golshan, a senior director of AI software at Nvidia, told China Daily at GTC when asked about how he viewed the "raising lobsters" phenomenon in China. "I think from our perspective, it's been phenomenal, because it's generating these very valuable use cases in the community," he said.

He said this kind of moment is similar to what happened when the web was created, when people suddenly found remarkably creative ways to build entire businesses, new products, and fresh ways to reach an audience.

Nvidia CEO Jensen Huang told an open models panel on Wednesday that "in a lot of ways, OpenClaw is bringing agentic systems to the consumer mindset". Harrison Chase, co-founder and CEO of Lang-Chain, echoed, saying at the same panel that OpenClaw represents a transition in what AI can be.

OpenClaw, Chase said, exemplifies a shift that began with professional software tools last year and is now reaching a larger population, which has transformed AI from something that simply responds to queries to something that takes real-world action on a user's behalf.

He predicted that this year would bring a new wave of personal productivity agents capable of autonomously handling more complex, longer-running tasks.

Yet the excitement has been tempered by serious caution. An earlier panel on AI agents in enterprise software drew broad agreement among participants that the technology's expanding capabilities demand equally robust safeguards.

Elia Zaitsev, chief technology officer of cybersecurity firm CrowdStrike, described AI agents as "a devil's bargain". The greater the power the technology offers and the more use cases it can solve, the greater the risks it introduces, he said.

That tension between capability and risk has already drawn regulatory attention in China. The Chinese authorities have issued several alerts warning that OpenClaw could expose organizations and individuals to significant cybersecurity vulnerabilities. In response, Tencent Cloud on Wednesday unveiled an upgraded enterprise-grade solution aimed at making AI agent deployment safer and more scalable.

Nvidia also unveiled "Nemo-Claw" at GTC, an open-source stack designed to layer privacy and security controls onto OpenClaw, a direct response to the growing concerns surrounding the technology.

Golshan of Nvidia framed the security challenge in terms of a maturity journey, comparing the road ahead for agentic AI to the path once traveled by the web and internet browsers. That journey, he said, ultimately requires the construction of a trust layer, which means focusing on the privacy and security controls that can help developers get started safely.

"I think any technology has to be smart and make sure it's secured. So there's nothing wrong about making sure that you're putting some kind of governance," Amit Zavery, president and chief product officer of ServiceNow, told China Daily.

He said organizations must remain alert to technological transformation and take responsibility for managing both data and security. "As long as you are making sure it's secured and putting some governance, it's a valuable thing as part of the enterprise," he said.