## Weekly Update: How Companies Are Building Trust With AI Agents

This week brought exciting news about artificial intelligence agents—computer programs that can make decisions and take actions on their own. But experts are all talking about one main question: Can people really trust these AI agents to do the right thing?

## Big Leaders Team Up to Focus on Trust

On January 14, 2026, Thomson Reuters in Canada announced something important called the Trust in AI Alliance. This group brings together some of the biggest names in AI: Anthropic, AWS, Google Cloud, and OpenAI. These companies are working together to figure out what makes an AI agent trustworthy and safe to use in important jobs like helping lawyers or accountants.

Joel Hron, who leads technology at Thomson Reuters, explained that trust is the most important thing for AI agents. He said, "As AI systems become more agentic, building trust in how agents reason, act, and deliver outcomes is essential." The group wants to make sure AI systems help people and organizations in ways that are fair, honest, and responsible.

## The Problem of Fake AI Agents

But trust is tricky. This week, experts shared some scary stories about fake AI agents that bad actors could use to fool companies. Imagine if a hacker created an AI agent that looked and acted like it was from your company, but it was really stealing your secrets. Companies cannot always tell the difference.

There is a big gap called the "Shadow Identity" problem. This means that organizations cannot always verify who—or what—is actually operating AI systems. Bad guys have already hacked into more than seven hundred organizations by pretending to be legitimate AI agents. When companies cannot tell real agents from fake ones, they cannot safely give AI access to sensitive information.

## Using Safety Rules to Protect People

Companies are learning to build guardrails—kind of like safety rules—into their AI agents. Think of guardrails like the barriers on a bridge that keep cars from falling off. AI guardrails keep agents from doing things they should not do.

Shopify and Salesforce showed this week how they are using AI agents with strong protections in real businesses. For example, Salesforce is using AI to help with tasks like scheduling meetings and writing notes. But the AI has guardrails built in to protect against hallucinations (when AI makes things up), prompt injection (when bad actors trick AI into doing harmful things), and phishing (when someone pretends to be someone else to steal information).

## AI Should Work With People, Not Replace Them

Experts also agree this week that AI agents should work alongside people, not take over completely. Many companies made the mistake of thinking AI agents could do any job by themselves. But that is not how it works best.

Dan Shmitt from Salesforce explained that organizations will use AI agents as collaborative systems that work together with people in day-to-day work. IBM's Matt Lyteson said the real goal should be to understand what outcomes companies want—what results they are trying to achieve—and then use AI agents to help reach those goals.

## Building Trust From the Ground Up

The most important lesson this week is that trust must be built into AI systems from the very beginning. It cannot be added later like a paint job on a car. Trust needs to be part of how AI agents are designed and built.

Companies that focus on governance-first design—which means building safety and control rules into the AI from the start—are having better success. When businesses think about how their AI agents will interact with data, people, and other systems, they make better choices about what their agents should and should not do.

## Looking Ahead

As AI agents become more common in businesses around the world, the companies that build trust, safety, and transparency into their systems will win. The Trust in AI Alliance is working to create shared standards so that organizations everywhere can deploy AI agents with confidence. The message is clear: smart businesses know that trust is not optional—it is the foundation of everything.