TM Forum names Infosys team winner of AI trust hackathon

TM Forum said more than 200 developers across 40 teams tested practical ways to prove trusted autonomous AI during a three-day hackathon in Dallas. The winning Infosys team showed how governance controls can be enforced at runtime without slowing AI deployment.

Why it matters: - TM Forum is trying to turn trustworthy AI from a policy goal into something operators can prove in live systems. - The work matters for telecom and other enterprise networks that want more autonomous AI without losing control, auditability or compliance. - TM Forum said global customers, regulators and partner ecosystems need a common approach, not separate trust models from each operator.

What happened: - TM Forum named Infyvibes, a team from Infosys, the winner of its Trustworthy AI & Data Hackathon. - The hackathon ran during Innovate Americas on Oct. 6-7 in Dallas. - More than 200 developers across 40 teams took part, both onsite and virtually. - Participants worked in a live autonomous enterprise sandbox built on TM Forum’s AI-Native Open Digital Architecture Canvas.

The details: - Developers configured agents and tools, tested governance controls and pushed the environment under sustained, and sometimes hostile, activity. - The challenge focused on proving that agentic decisions can be controlled, explained and verified without slowing deployment. - Infyvibes turned governance requirements into verifiable operational assurance. - The team showed how autonomous AI operations can be stopped and routed for human approval when defined limits are breached. - Infyvibes also demonstrated real-time enforcement and human-in-the-loop escalation to preserve accountability and control. - The team met all three required controls in the challenge: spend, event recording and performance drift. - Across the event, participants created 196 additional agents and executed more than 35,000 governed model calls. - Teams used 50 running ODA Canvas instances, along with Canvas documentation and repositories, to build working solutions. - TM Forum said its Model-as-a-Service gateway governed model traffic during the hackathon by applying policy, identity, authentication, encryption and guardrails in AWS. - Other teams built an intrusion-detection capability for agents from infrastructure audit information and tied it to a compliance dashboard. - The hackathon was supported by Accenture, AWS, Dell Technologies, NVIDIA and ServiceNow. - AT&T hosted the event.

Between the lines: - TM Forum is treating trust as an engineering problem, not just a compliance exercise. - The scale of the event suggests the organization wants developers to learn a shared operating model for AI-native systems, not just a one-off demo. - The runtime controls and human escalation paths point to a broader industry need: more autonomy, but only if operators can intervene when risk rises.

What's next: - TM Forum said it will take the code, evidence, controls and lessons from Dallas to its global membership. - The organization plans to turn the hackathon output into common methods, standards and industry assets for governing AI-native operations. - TM Forum’s next hackathon is set for Accelerate 2027, Feb. 1-4 in Amsterdam. - More information is available at TM Forum.