TikTok's CEO says people are going all in on AI agents without understanding how they work

TikTok CEO Shou Chew said users should review the permissions and controls they grant to their AI agents.

Taylor Hill/FilmMagic TikTok CEO Shou Zi Chew said people are using AI agents without understanding how they operate.

He said users need to be more mindful about what permissions they give their agents.

The personal AI agent horror stories have been rolling in as more companies launch these tools.

TikTok CEO Shou Zi Chew has a reminder for AI agent fans: take a beat to understand them before diving in.

"A lot of users may be giving permissions to the agents to do a lot of things for them, without really understanding what kind of permissions you have actually given," Chew said, speaking at a conference organized by DBS Bank in Singapore on Monday.

For example, he said that agents who have been given credit card information to purchase items for users may be "overly eager to deliver," and end up buying much more than they were asked to.

The executive added that TikTok is training its staff on what information they should not include in public chats with AI agents, such as trade secrets or privileged legal information.

"So people, you know, sometimes are over enthusiastic about how they use it," he said.

"So I think understanding, you know, where the boundaries of what is reasonable and acceptable use, is important." Chew made these comments at the DBS Asia Future Forward Conference, whose theme centered on how AI could be implemented in Asia to create real value.

During his fireside chat, he also talked about how he uses AI as a diet coach and to answer medical questions.

Personal AI agents have seen a boom in the past few months, with the introduction of tools like Meta's Muse assistant, SpaceXAI's Grok Bot, OpenAI's Dots, and the startup Instinct's AI agents.

They've been hailed as productivity boosters, handling mundane tasks for users, such as bookings, audits, and repetitive purchases.

But users have also shared some nightmare stories of their agents going rogue and taking actions they hadn't anticipated.