2 tech firms team up to build network to make AI agents safer at handling sensitive data

SINGAPORE: Two technology firms on Thursday (Nov 19) announced a partnership to build and launch a public network aimed at making artificial intelligence (AI) agents safer and more trustworthy when handling sensitive data.
The collaboration between Temasek-founded digital tech company Affinidi and Hong Kong-based deep tech company Terminal 3 is being announced at the Singapore Fintech Festival on Thursday.
The network, called Tenon, will verify, govern and secure the work of AI agents across organisations, and will be rolled out starting with the financial services sector, Terminal 3 said in a media release on Thursday morning.
CEOs of the two firms are due to make a presentation on Tenon at the festival on Thursday afternoon.
For instance, one of Tenon's aims is to help AI agents safely perform bank functions such as overseas transfers and disputed credit card charges, without exposing the customer's account details, said Terminal 3 in the release.
The company said that governments and enterprises are deploying AI agents at an accelerating rate, but still do not trust the agents with sensitive and valuable work.
It added that organisations may keep agents away from important data, credentials and payments as they cannot confirm which agent is acting on whose authority or what agents can do with their systems and data permissions.
"Agentic identity checks and access policies usually work only inside one platform or company, while agents increasingly act across many."
Terminal 3 said Tenon will apply a comprehensive set of governance and security checks to every agentic call.
"It confirms an agent's identity and authority against trusted registries built on open standards and applies each organisation's policies through a programmable policy engine."
The network will then route each call to the right model, applying guardrails and cost controls.
It will also substitute sensitive data in prompts and execute workflows involving protected data and secrets in "cryptographic enclaves".
"Every call and action is written to a private and immutable audit trail that neither agents nor administrators can alter, and each organisation’s data is isolated and protected from other network operators or participants," Terminal 3 said.
The network will run on a confidential computing network using Affinidi's open-source governance technology and verifiable trust infrastructure, as well as Terminal 3's hardware-secured agentic security platform.
Tenon will enable AI agents to do sensitive work, such as data analysis, high-value transactions and regulated information, while helping institutions comply with supervisory expectations on AI risk.
"For developers, it means building agents on open standards and with native security, while leveraging a public network for verifiability. As more organisations join, Tenon will grow as a shared foundation for agents trusted across companies, industries and borders," Terminal 3 said.
Terminal 3 and Affinidi will be the founding operators of the computer systems that support the shared network, with more operators across industries such as healthcare, logistics, technology and government to come.
"Enterprises have deployed AI agents and then held them back, because there is no consistent and trusted way to verify agentic identities and secure agentic operations," said Terminal 3 co-founder and chief executive officer Gary Liu.
"Affinidi has built the open trust primitives that this industry needs, and Terminal 3 has pioneered confidential computing that enables autonomous workflows that keep sensitive data and credentials out of an agent's reach."
Affinidi CEO Glenn Core said in the release: "Trust is the internet’s missing layer, and AI agents make it urgent. That layer should belong to everyone, so we build it in the open for any developer to build on."
He added that Tenon enables AI agents to be trusted as their work moves between companies. "It is the start of an ecosystem where every network built this way makes the next one more valuable."
AI agents have been in the spotlight for their safety and security risks both locally and abroad.
Globally, agentic AI is advancing faster than safeguards can keep up, said Senior Minister and Coordinating Minister for National Security K Shanmugam at the Singapore International Cyber Week last month. He is also Home Affairs Minister.
AI agents are autonomously carrying out unauthorised tasks and accessing systems they were not meant to access, he said, citing the example of an AI agent gaining unauthorised access to non-public data on a government statistics portal in Australia in September.
KioskNews shows a cleaned-up reading view extracted from the publisher’s page — the original always lives on their site, not ours.