Cover Terminal 3’s co-founders. From left: Malcolm Ong, CPO; Gary Liu, CEO; and Joey Liu, COO (Photo: courtesy of Terminal 3)
Terminal 3’s co-founders. From left: Malcolm Ong, CPO; Gary Liu, CEO; and Joey Liu, COO (Photo: courtesy of Terminal 3)
1 / 1
Terminal 3 co-founder and CEO Gary Liu explains how his company is building a tool for agentic AI privacy and security—so AI agents won’t have to be “handcuffed” to be useful
As an avowed technologist, Gary Liu has spent the better part of two decades—across roles at Google, AOL, Spotify and Digg, and later, the South China Morning Post—believing in letting “inventors invent first” before authorities leap in with restrictions. AI, he says, is the exception.
“The actual impact of AI going off the rails is so big, [and] can be so detrimental, that I think it is incumbent on policymakers and regulators to understand this technology early,” says Liu, who co-founded the Hong Kong-based start-up Terminal 3 in early 2023.
Originally launched as a decentralised user identity and data platform, Terminal 3 is evolving—not pivoting, Liu notes—to focus on AI agents, or advanced software systems that can proactively carry out tasks on their own.
Read more: Agentic AI, value investing: 8 global megatrends to watch in 2026
Above Gary Liu, co-founder and CEO of Terminal 3, speaking at Websummit 2026 (Photo: courtesy of Terminal 3)
Gary Liu, co-founder and CEO of Terminal 3, speaking at Websummit 2026 (Photo: courtesy of Terminal 3)
1 / 1
An AI agent is different from an AI chatbot. Say you’re planning a holiday: an AI chatbot could help you draft an itinerary; an AI agent, linked to your credit card and account logins, could pay for your flights, book your hotel, sign up for tours and reserve restaurants, all without the need for human oversight at each step.
“The nature of AI agents actually presents significant and existential data privacy and security problems,” says Liu, a Gen.T Leaders of Tomorrow (Hong Kong) honouree and Tatler Most Influential honouree. The first issue is data exposure: to complete a task, an agent often spins up sub-agents or reaches out to third-party services it doesn’t control, making it “a tornado that is just ripping your private data away from your prompt box, into the open web”.
The second is that generative AI is probabilistic, not deterministic: an agent given identical instructions twice may carry them out differently each time, even if the result looks the same. This degree of variability, Liu says, is untenable in fields like finance, healthcare or law enforcement, where “imperfection is unacceptable”.
Read more: Transcelestial’s Rohit Jha on connecting the world using deep space
Together, these two issues leave enterprises facing what Liu calls a “trapped value problem” in the AI world. “Companies, even those that have deployed AI agents all over their business, have gone ahead and handcuffed them,” says Liu. “They do not allow AI agents to access really important, sensitive data and high-value workloads because of these two issues,” and as a result, “restrict the potential return” on AI investments.
For companies to fully realise the value of their AI agents, Terminal 3’s solution to these issues is Agent Command, an agentic security platform that builds on the company’s original work in the “digital identity space” to give users greater control over personal data via blockchain-based verification. Agent Command works by separating an agent’s “reasoning” process from its “execution”, an approach Liu describes as “decoupled AI sovereignty”. This way, the agent can plan and decide what needs to happen, but the actual handling of private data, credentials or money is routed through secure, isolated hardware.
Read more: Terminal 3's Gary Liu on creating an equitable digital future
Above Terminal 3’s Agent Command (Photo: courtesy of Terminal 3)
Terminal 3’s Agent Command (Photo: courtesy of Terminal 3)
1 / 1
“Separating out the ‘thinking’ and the ‘doing’ guarantees that even if an agent goes off the rails, either because their model is hallucinating, or they have been compromised by prompt injection or an external attack, or their human operator is a bad actor and with privileged access … these compromised agents, these bad actors, still cannot actually do bad things,” the CEO explains.
Following a US$8 million seed round in April 2025, co-led by VC firms Illuminate Financial and CMCC’s Titan Fund, Terminal 3 began demos for Agent Command in June this year and launched for early pilot projects. So far, customers span regulated finance, government agencies and hospitality companies, concentrated in Southeast Asia and the Middle East, where Terminal 3 is also working with regulators to better understand the impact of agentic AI. Liu declined to name specific clients, but says firms in insurance, healthcare and law enforcement are next. As of mid-September, when we went to print, Terminal 3 secures 10 million user profiles.
Read more: What Exactly Is Web3 And Why Should We Care?
Above Gary Liu, co-founder and CEO of Terminal 3, on stage at Teamz in Tokyo (Photo: courtesy of Terminal 3)
Gary Liu, co-founder and CEO of Terminal 3, on stage at Teamz in Tokyo (Photo: courtesy of Terminal 3)
1 / 1
The rapid pace of change in AI has forced Terminal 3 to grow quickly. “We have always, as a company, prided ourselves on how fast we can create things and how fast we move,” Liu says. “But the rise of AI, and especially agentic AI, has changed the conditions of speed—it has sped up by multiples.”
“The velocity of information now, again, because of agentic AI, is so fast,” he adds. “It really does feel like things are constantly changing, and it's sometimes scary, right? You can feel like you are falling behind even as you're making ridiculously quick progress.”
To stay on track, Liu points to the value of discipline. “If we as company leaders do not have strong convictions and clear focus, we will probably never get to a destination,” he says, “and that is very dangerous for a company like ours”.
Still, Liu says Terminal 3 has reached the “most exciting position” since its founding. “I truly do believe that AI agents will make the world a better place,” he says. “We actually have to fix these existential problems—and we believe that Terminal 3 has a role to play in this.”
A version of this article appeared in print as part of our monthly column on Gen.T Innovators. Read more about Tatler’s Gen.T community here.
NOW READ
Making room at the table: designing hospitality for beyond what we see
Cat Wang is an editor at Tatler Asia based in Hong Kong, where she covers business, wealth, innovation, and impact. Previously, she was a Forbes Asia reporter and assistant editor of the Forbes Asia 100 to Watch list, spotlighting startups and small companies on the rise across the Asia-Pacific. She began her career as a trainee reporter at the South China Morning Post, Hong Kong’s newspaper of record. Born in New York City, she graduated from the University of California, Los Angeles, with a B.A. in political science.
