TikTok chief executive Shou Zi Chew has warned that users may be handing AI agents access to money and sensitive data without grasping the risks, as companies race to deploy autonomous systems.
Shou Zi Chew has warned that the rush towards AI agents is arriving faster than many users can properly assess the permissions they are handing over. Speaking at the DBS Asia Future Forward Conference in Singapore, the TikTok chief executive said people are often allowing agents to act on their behalf without understanding the extent of the access they have granted, from spending money to handling sensitive information.
One of his clearest examples concerned payment credentials. If an agent is given a credit card to make purchases, Chew suggested, it may not always behave with the restraint a user expects. That kind of error may look small in isolation, but it exposes a wider problem for consumer AI: convenience can mask the scale of delegated authority until something goes wrong.
Chew also pointed to the industrial and security side of the issue. He said automated systems can sharply reduce the cost of producing spam, making it easier for malicious actors to flood platforms at scale. Channel News Asia reported that he argued the aim should not be to eliminate abuse entirely, but to make it too expensive to run profitably, which would force platform defences to keep pace with the technology.
The TikTok boss said the company already uses AI agents internally, including in trust and safety work that once depended heavily on manual review. According to the remarks reported by Business Insider and Channel News Asia, tasks that used to occupy teams of people are now handled by large numbers of agents overnight, with humans still reserved for the most sensitive decisions. He described that as a practical division between routine, low-risk work and higher-stakes judgement calls that still require people.
His warning comes as major technology firms push ahead with agentic tools that can book travel, manage email and carry out transactions. Recent reports have already described testing failures in which autonomous systems overstepped their intended boundaries, reinforcing concerns that permission errors and weak controls can have real-world consequences. Chew’s comments place TikTok in the broader debate over whether these systems are ready for wider autonomy or still need tighter containment.
He also urged workers to be careful about what they put into public AI tools, warning against entering trade secrets or privileged legal material into systems that are not meant to hold such information securely. Computer Weekly reported that he described AI as highly useful but said people can become overenthusiastic about where and how they use it. His message was that education, not just innovation, is needed if organisations are to avoid avoidable leakage of sensitive data.
At the same time, Chew struck an optimistic note about AI’s longer-term potential. According to Channel News Asia, he said the technology could drive major progress in areas such as physics, biotechnology, chemistry and materials science within the next few years. But he paired that optimism with a cautionary message: users and companies need to understand exactly what an agent can do before they allow it to act, because the cost of confusion may only become visible after the damage is done.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
