Marketing Leadership · · 4 min read
What a CEO Should Ask About AI Risk
By Scott Hashisaki, Fractional CMO & Growth Executive
Google's latest threat report shows attackers using AI agents to steal credentials in hours. Seven questions every CEO should put to their team, vendors and agency.
Most AI risk conversations in the boardroom are about the wrong thing. They focus on whether AI will write a bad email or invent a fact. Google's Q3 2026 AI Threat Tracker points somewhere more practical: the accounts, keys and connections that AI tools run on.
In one case Google's Mandiant team investigated, an attacker got into a company's cloud environment and used an AI coding assistant plus a set of written agent instructions to build and run a mass credential-harvesting operation in under six hours. Google also noted that attackers are going after AI assets directly: API keys, cloud compute, and the configuration files that tell AI agents what to do. And it was careful to say fully autonomous attacks at scale haven't shown up in the wild yet.
For a CEO, the takeaway is simple. Your company's AI footprint is now part of your attack surface, and most of it was added quickly, by different people, without a single owner. Here are the questions I'd ask this week.
1. What AI tools are we actually using?
Not the approved list, the real one. Chatbots on the website, AI features inside the CRM, writing tools, meeting recorders, coding assistants, agency-run automations. Marketing and sales are usually where this sprawls fastest.
2. What can each tool connect to, and can it change things or just read?
An AI tool with read access to a calendar is a different risk from one that can send email, edit the website or move money. Ask for the connection list in plain language.
3. Where do the credentials live, and when were they last rotated?
Every connection runs on a key or login. Who holds them, where are they stored, and what happens to them when someone leaves?
4. What are our vendors and agencies doing with AI on our behalf?
If an outside partner runs AI tools inside your accounts, you're carrying part of their risk. Ask what tools they use, what access those tools have, and how they'd tell you about an incident.
5. Who owns this?
Not IT in general. A named person who keeps the inventory current and has the authority to shut something off.
6. How fast would we know?
Google's six-hour example is the point. If your answer to "how would we notice?" is "a customer would tell us," that's the gap to close: alerts on new admin accounts, unusual logins and unexpected AI usage costs.
7. What's our off switch?
If a tool or a key is compromised, can you revoke it in minutes? Write that plan down before you need it.
Why this sits with leadership
None of these questions requires a security degree. They require someone to ask them and someone to own the answers. AI is genuinely making lean teams faster, and that's worth protecting. The companies that get the most from it will be the ones that know exactly what their tools can touch.
If your marketing and growth stack has grown faster than anyone can map, that's a conversation I have with leadership teams regularly. [Let's talk](/contact).
Source: Google Threat Intelligence Group, AI Threat Tracker Q3 2026, as reported by [Help Net Security](https://www.helpnetsecurity.com/?p=383967) (September 2026).