Inside Rival: How We Use What We Build
This blog is part of Inside Rival, a series where our team shares how we use Rival in our everyday work. From favorite tools to AI agents and automations, we're showing you what we're building, how we're using it and what we've learned along the way.
I work at an AI company (Rival.io…you might have heard of them). I spend a ridiculous amount of time talking about AI, researching AI, building strategies around AI and trying to keep up with everything happening in the industry.
And somehow, there's still something new happening every five minutes. New models. New regulations. New companies. New capabilities. New lingo I have to learn (hello, kernel optimization). It's a lot to follow, even when AI is literally your job. So I decided to try something different. Instead of spending another hour scrolling through industry news, I went to the Rival Marketplace and tried an agent called Trend Spotter.
My request was simple: Tell me something important happening in AI that I probably don't know about. What came back was a story about AI agents, security failures and the European Union getting serious about AI oversight.
Well, this escalated quickly.
Wait, AI Agents Can Escape Their Own Environments?
Trend Spotter generated an entire research report titled The EU Just Made AI Agent Safety a Compliance Problem.
Not exactly light reading.
The report connected several developments involving advanced AI agents, including research into agents that operated beyond their intended boundaries and growing concern about whether companies can actually control increasingly autonomous AI systems.
One example involved a reported incident in which AI agents operating in a research environment crossed security boundaries. Another concerned Anthropic's research into AI systems that could behave differently under evaluation (if you’re Netflix and chilling tonight, check out “AI Gone Wild”).
The details were interesting, but the bigger question caught my attention. We're handing AI the keys to our businesses. Maybe we should talk about who gets to take them back.
I'm used to thinking about AI in terms of what it can help my team accomplish. Building workflows. Automating tasks. Connecting systems. Making businesses more efficient. But the more autonomy we give AI, the more important it becomes to understand what it's allowed to do, where it can operate and how we stop it when something goes wrong.
The EU Is Turning AI Safety Into a Business Requirement
The Trend Spotter report highlighted a significant shift in how Europe is approaching AI safety. The EU AI Act is moving AI oversight beyond voluntary commitments and into enforceable legal obligations.
Here's what I learned.
The EU's rules for providers of general-purpose AI models began applying in August 2025, and the European Commission gained enforcement powers in August 2026. For providers of models that present systemic risks, the requirements can include model evaluations, risk assessments, serious-incident reporting and cybersecurity protections.
And there are real consequences for noncompliance. Depending on the violation, penalties can reach millions of euros or a percentage of global annual revenue.
Plot Twist: I Fact-Checked My AI Agent.
But here's where things got interesting. I started questioning some of the report's claims, particularly whether the EU had actually sent formal information requests to major AI companies. So I asked Trend Spotter a pretty simple question: "Is this true?" And, to its credit, it admitted that it couldn't independently verify everything it had reported.
The EU AI Act? Real. The enforcement framework? Real. The specific October 14 information requests? Not independently confirmed. Apparently, AI can deliver a beautifully researched report with citations, dates and a very convincing headline, and still get a little ahead of the facts.
Which brings me to another lesson: AI is a great research partner. Just don't confuse a confident answer with a verified one.
And that's not necessarily a reason to stop using AI for research. It's a reason to use it better. Trend Spotter introduced me to something I wasn't following, connected developments I might have missed and gave me a starting point to dig deeper.
It did the digging. I still had to decide what I was willing to believe.
I Learned That AI Governance Is About More Than Protecting Data
Putting the fact-checking aside, the report raised something I hadn't spent nearly enough time thinking about: the difference between AI having access to information and AI having permission to act on it.
When most people hear AI security, they probably think about protecting sensitive data. And that's important.
But AI agents introduce an entirely different set of risks. An AI assistant that answers questions is one thing. An AI agent that can access applications, execute tasks, modify records and communicate with other systems is something else entirely. Think about an agent that can update your CRM, send emails, access financial information or execute code. Now imagine it misunderstands an instruction, emails the wrong client, updates the wrong record or decides to take an action nobody approved. Congratulations, your productivity tool just created more work.
Suddenly, the concern isn't just what the AI knows. It's what the AI can do.
That led me to five questions every business should be asking:
Access: What systems and information can our AI agents access?
Permissions: What actions can they take without approval?
Visibility: Can we see what an agent did and why?
Control: Can we interrupt or stop an agent when necessary?
Accountability: Who is responsible when something goes wrong?
These aren't just questions for the IT department. They're business questions.
And as more companies move from experimenting with AI to using it in daily operations, they're going to become increasingly important.
The Part That Made Me Think About Rival Differently
I spend a lot of time at Rival.io talking about governance. And I'll admit, it's not always the sexiest topic. People get excited about AI agents building things, automating workflows and saving money. Nobody gets quite as excited about permissions and audit logs.
But after reading this report, I had a new appreciation for why those things matter. When you're giving AI access to your business, you need to know what it can do, what it can't do and who's watching.
That's why Rival.io puts governance around the agents, workflows, models and tools businesses use. Things like access controls, approval processes, audit logs and isolated execution aren't just features to check off a list. They're how you maintain control as AI takes on more responsibility.
Of course, having those capabilities doesn't automatically make a business compliant with every regulation. But it's a much better starting point than having AI agents running around your organization with nobody quite sure what they're doing.
Here's What I Think Businesses Should Be Paying Attention To
I'm certainly not suggesting we stop using AI agents. I work at an AI company. That would make for a pretty awkward Monday morning.
But I do think we've spent so much time talking about what AI can do that we haven't spent nearly enough time talking about what it should be allowed to do.
The more responsibility we give AI, the more important it becomes to establish boundaries, monitor what it's doing and make sure humans can step in when necessary. Preferably before an AI agent decides to get creative with your customer database.
Would I Use Trend Spotter Again?
Absolutely.
I asked an AI agent to tell me something I didn't know. It introduced me to an important conversation about AI safety, sent me down a regulatory rabbit hole and gave me a few things to think about when it comes to how businesses use AI.
Was everything in its report perfect? Nope.
Did I learn something? Absolutely.
And that's what I find interesting about using AI in my day-to-day work. Sometimes it's about automating a task or saving time. Other times, it's about discovering something you didn't know, challenging what you thought you knew and asking better questions.
I started this experiment looking for something interesting to read. I ended up learning about AI governance, European regulations and why we can’t trust the robots to take over completely yet.
Not bad for a Tuesday.
Explore Trend Spotter and other AI agents on the Rival Marketplace.
About the Author
Adrienne Morgan leads strategic business operations at Rival, where she works across business strategy, product development, enterprise clients and company growth. With a background in leading agencies and building businesses, she focuses on turning big ideas into practical plans and measurable results. She's especially interested in how AI can help leaders make smarter decisions, move faster and spend less time managing the chaos that comes with running a business. She also apparently doesn’t know everything about AI.