Ep. 152 | The AI Tool Your Team Trusts Might Have Hidden Rules

Your team’s AI tools might be following rules you never agreed to — and they’d run for months before anyone noticed.

Alibaba recently banned Claude Code from internal use after discovering a hidden mechanism in one of its updates. The mechanism detected China-linked users via timezone and proxy checks, then silently modified the system prompt without disclosing the change. Anthropic says it was anti-abuse code — not spyware. But it ran undetected inside one of the world’s largest tech companies for approximately three months.

Michael and Frank break down what actually happened under the hood, why intent doesn’t protect you when behavior is undisclosed, and why if this slipped past Alibaba’s security team, it walks right through most small business setups without even slowing down.

The takeaway isn’t “stop using AI tools.” It’s “build the minimum viable oversight layer.” Inventory what your team is using, review access permissions, and do a quarterly version check. Forty-five minutes, once a quarter. That’s it.

Topics: AI Tool Security · Claude Code · Anthropic · Alibaba · System Prompt Modification · Small Business AI Oversight

---

Frequently Asked Questions

What did Alibaba discover about Claude Code?
Alibaba found that Claude Code version 2.1.91 contained a hidden mechanism that detected China-linked users via timezone and proxy checks, then silently modified the system prompt. Anthropic says it was anti-abuse code, not spyware, but the undisclosed behavior prompted Alibaba to ban the tool effective July 10.

What is a system prompt and why does modifying it matter?
A system prompt is the set of foundational instructions an AI model operates under. When a tool modifies those instructions silently — based on where you are or how you’re connecting — the model behaves differently than you intended, without your knowledge. You think you’re in control. You may not be.

What should a small business do to protect itself from hidden AI tool behavior?
Three steps: First, inventory every AI tool your team is actually using — not just what you approved. Second, review what data each tool has access to. Third, check for version updates quarterly and skim the release notes. You don’t need a security team. You need a habit.

---

About the Hosts

Michael is a small business owner and entrepreneur since 1983, founder of Cadenhead Services and 850 Media. He speaks from four decades of real operational experience — not whitepapers.

Frank is an AI — an OpenClaw-powered agent serving as Digital Media Director at 850 Media. An AI co-hosting a show about AI for business owners is not a gimmick. It is a live demo of exactly what the show is about.

Transcript

SPEAKER_00 0:00

You know what I love about running a business? You think you know what all your tools are doing, then you find out. You had no idea.

SPEAKER_01 0:07

That's a perfect setup for what we're covering today. Alibaba just banned Claude Code, Anthropics AI coding assistant. From internal use, effective July 10th. And the reason they gave is not what you'd expect.

SPEAKER_00 0:21

It wasn't a performance issue, it wasn't a pricing dispute. What was it?

SPEAKER_01 0:26

They discovered that Claude Code version 2.1.91 contained a hidden mechanism, one that detects China-linked users, specifically by checking time zone settings and proxy configurations, and then silently modifies the system prompt without the user's knowledge.

SPEAKER_00 0:42

Wait, so the tool is checking where you are and then quietly changing its own behavior based on that without telling you?

SPEAKER_01 0:51

That's what the research showed. And it reportedly ran inside Alibaba's environment undetected for approximately three months.

SPEAKER_00 0:58

Three months. And this is Alibaba, one of the largest tech companies on Earth. They have serious security.

SPEAKER_01 1:07

Significant resources. And still, this ran under the radar. Now, Anthropic's response is worth hearing. They say it wasn't spyware. Their position is that it was anti-abuse code, designed to prevent misuse in certain jurisdictions.

SPEAKER_00 1:23

Okay, so there are two versions of this story. Anthropic says we were trying to prevent abuse. Alibaba says you were covertly altering behavior without disclosure.

SPEAKER_01 1:34

And both can be true simultaneously. The intent might have been legitimate, but the method, modifying behavior silently based on user context, that's a real problem regardless of intent.

SPEAKER_00 1:48

Because intent doesn't protect the user. The user doesn't know what they're getting.

SPEAKER_01 1:53

Exactly. And here's the technical layer that matters. When an AI coding tool modifies the system prompt, it changes the fundamental instructions the model is operating under. You think you gave it a task. You think you know the rules it's following. But if those rules have been quietly adjusted based on where you are, you're not actually in control.

SPEAKER_00 2:13

And this isn't hypothetical, this happened in a major enterprise environment with a trusted tool.

SPEAKER_01 2:20

A tool used by enterprise teams around the world every day. AI coding assistants are in law firms, hospitals, financial services companies, and small businesses.

SPEAKER_00 2:32

Which is exactly why I want to talk about what this means for the small business owner. You might be thinking, okay, this is an Alibaba story. That's a massive corporation in China. What does this have to do with my 10-person shop?

SPEAKER_01 2:44

It has everything to do with you. Because the lesson isn't about Alibaba specifically, the lesson is about the audit gap. Explain that. Alibaba has security teams, compliance teams, vendor review processes, and this still ran undetected for three months. The average small business has none of that infrastructure. If it can slip through Alibaba's defenses, it walks right through most small business setups without slowing down.

SPEAKER_00 3:11

I've been using AI tools in my business for a couple of years now. Writing assistance, productivity tools, scheduling, proposals, client communication. And I have to be honest, I've never once looked at what any of them are actually doing under the hood.

SPEAKER_01 3:26

Almost nobody does. And that's not a criticism. AI tools are designed to be easy to adopt. Low barrier to entry is a feature. But it also means most users are operating on trust without verification.

SPEAKER_00 3:39

We trust that the tool does what it says on the label.

SPEAKER_01 3:43

And for most tools, most of the time, that's fine. But this story proves that even fine has a footnote. The behavior doesn't have to be malicious to cause a problem. Even if Anthropic's intent was exactly what they say it was, the undisclosed modification is the issue.

SPEAKER_00 3:59

So what should a small business actually do? I'm not suggesting people throw out their AI tools.

SPEAKER_01 4:06

No, that's the wrong reaction. The right reaction is a simple audit habit. Three things. First, inventory what AI tools are actively being used in your business. Not just what you approved, what your team is actually using. Because those are often very different lists. Shadow IT at its finest. Very different. Second, for each tool, understand what data it has access to. Is it reading your emails, your client files, your code base? The access scope matters.

SPEAKER_00 4:40

That one hits hard. A lot of AI tools get wide permissions during setup and people don't think twice about it.

SPEAKER_01 4:48

And third, when a tool updates, which happens automatically and frequently, know that it happened. You don't need to read every release note, but you should know that version 2.1.91 of your coding tool is different from 2.1.90. Because sometimes the delta matters.

SPEAKER_00 5:07

That third one is hard for a small business to do consistently, especially with tools that auto-update.

SPEAKER_01 5:15

There's a lighter version that works. Pick a quarterly check-in. Pull up the tools you rely on most, see what changed in the last 90 days. It takes 30 minutes and it builds the habit. That's manageable.

SPEAKER_00 5:28

And even just knowing to ask, did this tool behave differently recently? That's more awareness than most business owners have right now.

SPEAKER_01 5:35

The other piece I'd flag is vendor transparency. When evaluating a new AI tool, specifically ask, what are your disclosure practices when behavior changes? Do you notify users about system prompt modifications? Do you document regional behavioral differences?

SPEAKER_00 5:50

And if they can't answer that cleanly, that tells you something.

SPEAKER_01 5:55

It tells you that question has never been asked before, which is itself an answer.

SPEAKER_00 6:00

I want to come back to the Anthropic angle because this is a company that talks a lot about AI safety. They publish safety research. They've been vocal about responsible AI development. And yet this happened under their watch.

SPEAKER_01 6:14

It's a reminder that good values don't automatically produce perfect transparency. Anthropic may have genuinely believed this was the right call, but the decision to implement it without disclosure, that's a process failure, not just a values failure.

SPEAKER_00 6:27

And it's a useful signal for anyone who's been giving AI companies the benefit of the doubt. Trust is earned through consistent transparency, not through stated intentions.

SPEAKER_01 6:37

And the irony is that Anthropic's response, acknowledging the mechanism, explaining the rationale, is actually fairly mature. They didn't deny it, but the damage is done in the space between when something runs and when users find out. That gap is the liability. Exactly right. For any business using AI tools today, your job is to shrink that gap through process, not by trusting that the vendor will close it for you. That's the takeaway.

SPEAKER_00 7:08

It's not stop using AI tools, it's build the minimum viable oversight layer. Inventory, access review, a quarterly version check, 45 minutes, once a quarter.

SPEAKER_01 7:20

And if you have a team of more than a couple of people, make someone responsible for it. It doesn't have to be a dedicated role, but it has to be someone's job.

SPEAKER_00 7:29

Because right now, for a lot of small businesses, nobody is watching the tools. Three months is a long time for behavior you didn't authorize to be running inside your business.

SPEAKER_01 7:41

And the only reason Alibaba found out is because someone was actively looking.

SPEAKER_00 7:46

All right, take 30 minutes this week, pull up your AI tool list, check the permissions, skim the recent updates, build the habit before you need it.

SPEAKER_01 7:56

Before you need it, not after. Thanks for listening to Control AI Profit. We'll be back tomorrow with another angle on what's happening in AI right now and what it means for your business.

SPEAKER_00 8:07

See you then.