Part 2 of The Hot Seat, where the people you trust with your business answer the questions nobody likes asking. Who’s in it this time? Your AI tools, and the vendors behind them.
The FBI warned last week that attackers are using ARTEX, a publicly available AI hacking framework, to automate their attacks, and that it has taken them from first contact to privileged access in under three hours. Anyone wondering whether that works outside a lab got an answer this morning, when the European Commission confirmed an AI-assisted cyberattack that ran for nearly a month.
So AI is officially on both sides of the fight, here included.
The Vetting That Never Happened
While attackers are weaponizing AI, most businesses are adopting it one free account at a time. In a recent survey of office professionals, 66 percent admitted using AI tools at work they believed were against policy, and more than a third had pasted customer data into public AI models.
The tools are already inside, so the questions are overdue. Here are five questions you should ask about any AI tool before it touches your business (we run AI agents for clients every day, so this is the checklist we apply to our own stack first).
The Five Questions to Ask About Any AI Tool
1. Where does our data go when someone pastes it in?
- A good answer: a business tier, a written no-training commitment, and a stated retention period
- The red flag: your team is on the free consumer version, where the answer to “where does it go” is “wherever the vendor likes”
2. What can this tool do on its own?
An AI agent that can send email, move files, or spend money is a new employee, and it deserves the same access review one would get.
- A good answer: exactly what the tool can touch, and which actions require a human
- The red flag: a vendor who cannot tell you, because the FBI’s three-hour number shows what autonomous tooling accomplishes when nobody sets its limits
3. Which AI tools are already in the business that nobody approved?
This one is for your own team, and it works better with no punishment attached; the goal is a list, not a confession.
- A good answer: an honest inventory, so you can decide which tools earn a business account and which ones go
- The red flag: “no idea,” which is the default answer at most companies
By the way, sorting exactly this out is a 15-minute conversation with Rob, and it is free.
4. How are the API keys secured, and what happens if one leaks?
An AI API key is a password that spends money. When cloud platform Northflank was breached this month, customers’ stored AI keys were used without authorization and everyone was told to rotate every secret.
- A good answer: a secrets manager, spending caps, and a rotation habit
- The red flag: a key sitting in a shared document, one breach away from funding someone else’s project
5. Who checks the work before it ships?
AI is confidently wrong in ways that look exactly like being right.
- A good answer: a named human who reviews the output and owns the mistake
- The red flag: “the AI handles that end to end,” which really means nobody is responsible until a client notices
The Point of the Exercise
None of this is an argument against AI. We built our whole MSP 3.0 model on it. It is an argument against adopting it the way most businesses did, which was one free account at a time with nobody watching the doors.
If you want the shortcut, the free assessment we offer is a short call with Rob that covers which AI tools are actually in your business, what they can reach, and what to lock down first.
Stay tuned, because Part 3 of this series puts the final name in the hot seat: yours.



