The Disclosure

Switch NovaAssist to human-only routing. Disable all chatbot auto-responses. Every query goes to a live agent.

That triples the queue. Three agents on shift until 1 PM.

Prioritise urgent queries. Anyone who asked about fees in the last 48 hours gets a callback today.

Amir, I need every piece of content you've used AI on. Blog, email, chatbot scripts, social. Last 11 months. By 11 AM. No judgment — just the facts.

Sorry. It's... basically everything. I'll get the list.

Good that you're auditing. But the chatbot is still live, and you just emailed 28,000 people. Every hour inaccurate content stays out there is additional exposure. An audit without action is just a paper trail of a problem you knew about.

I just read your email to the FT. You told her we have "human editorial oversight at every stage." Do we?

I—

If we don't, and she can prove it, that email is evidence of misleading a journalist. Article 50(4) — deployers must not misrepresent AI content. Article 99 — misleading info, up to EUR 7.5M or 1% of turnover. You've turned a transparency problem into a deception problem.

Elena, what's happening with the FT thing?

[Walks through journalist, chatbot, customer, scope of AI content]

How bad?

127 blog posts, 34 email campaigns, every chatbot script. Eleven months. No disclosure anywhere.

I'm back Thursday evening. Hold everything until then.

Journalist's deadline is Friday.

Tell her Thursday. Buy us a day. Don't tell her more than you have to — this isn't a confession, it's PR.

Niamh says it's compliance. Article 50—

The AI Act isn't fully enforced. We have until August.

Transparency provisions are already live for GPAI output. And the chatbot gave a customer wrong info. He's threatening the Central Bank.

Fine. Loop in Niamh. But "we used AI and didn't tell anyone" can't be the headline. Be smarter.

Fiona wants to manage the narrative. Here's what she's missing: Article 50 transparency on AI content is already enforceable for GPAI. And the chatbot — that's 50(1). Anyone interacting with AI must be told. Our chatbot discloses nothing. Not in the UI, not in the terms.

And Acheson?

The Air Canada precedent holds the airline responsible for what its chatbot said. Ours told him there was no maintenance fee. If we don't honour that, he's got a strong case — AI Act aside, this is basic consumer protection.

Mr. Acheson, my name is Elena Vasquez. I'm the Director of Marketing at NovaTech. I'm calling about your experience with our chatbot.

Finally, someone who isn't reading from a script.

Our chatbot gave you incorrect information about the premium fee waiver. That waiver was discontinued six months ago, but the chatbot wasn't updated. Our error, not yours.

So what are you going to do about it?

Three things. We're reversing the EUR 49 charge today. We're honouring the chatbot's offer — no maintenance fee for 12 months. And we're auditing every automated response for currency.

That's... fair. What I was asking for from the start. Your support team kept telling me the chatbot wasn't binding.

They shouldn't have said that. If our system gives you information and you act on it in good faith, we should stand behind it.

I appreciate the call. I'll take down the tweet.

The right call. The Air Canada tribunal held the airline liable because they tried to distance themselves from the chatbot. You did the opposite — defensible. If this reaches the Central Bank, we can show we identified, contacted, and remediated within 24 hours.

The journalist already knows the answer. If Fiona tells her Friday the article was human-reviewed, and Lena has evidence it wasn't, transparency becomes credibility. And on the customer — a refund without acknowledging the chatbot is AI doesn't satisfy Article 50(1). We fixed the charge, not the disclosure.

We've lost control of this. The journalist is publishing tomorrow with or without us. The customer has escalated to the Central Bank. Fiona isn't back until tonight. If she walks into the Friday call with "nothing to see here," this becomes the lead, not a sidebar.

What do we do?

We needed to do it yesterday. Article 50 required transparency from deployment. Every day we delay, the regulator's patience shrinks. I need to brief the board. This isn't a marketing problem.

You want me to tell the FT we used AI without disclosure for 11 months?

Tell her we found a gap, built a framework, and we're shipping it. The story is the framework, not the gap.

If the Central Bank opens an investigation — 60% chance, given Acheson's complaint — their first question is what we did on day one. A governance framework adopted within 72 hours is the difference between a warning letter and a formal investigation.

Cost?

Chatbot relaunch ~EUR 40k including review and redesign. Content audit: a contractor for two months, ~EUR 15k. Training internal. Under EUR 60,000 total.

Cost of not doing it?

Article 99: up to EUR 15M or 3% of global turnover — the deployer-obligations tier, not the lower misleading-info tier. For NovaTech, EUR 10.2M. AI Act alone — GDPR on the chatbot logs is separate.

Build the framework. I'll do the call.

Elena, I want to help. I should have asked about review. I should have checked the chatbot against the fee schedule. Can I help build the training?

Yes — you understand exactly what went wrong. That's not punishment, that's the most useful thing you can do.

This I can work with. Blog labels, review checklist, chatbot banner. We tell the journalist it's rolling out. Timeline?

Labels within a week. Chatbot banner within two.

Symptoms, not root cause. No documentation of which content was AI. No training. No process for chatbot verification. GDPR on the logs — untouched.

GDPR's separate. This gets us past Friday.

The Central Bank won't ask about banners. They'll ask if we have a risk management system, oversight processes, staff training. Article 9 requires it "established, implemented, documented, maintained." A checklist isn't a system.

Here's the plan. Tomorrow's journalist call — we position as leaders. "NovaTech is launching an AI transparency initiative." No chatbot. No 11 months without oversight.

Fiona, Niamh has flagged serious—

Niamh's doing her job. I'm doing mine. The board hears the narrative first, details later. Elena — tell your team, especially Amir, that nobody discusses our content processes externally. Not LinkedIn, not former colleagues.

Fiona, instructing staff not to discuss compliance concerns externally is whistleblower suppression. The EU Whistleblower Directive protects anyone reporting AI Act breaches. If Amir goes external and we've told him not to talk, our position is indefensible.

I'm not suppressing anything. I'm asking for message discipline.

A regulator won't see it that way.

Thank you, Fiona. To clarify — your chatbot, NovaAssist. Does it use generative AI?

It uses advanced natural language processing. For routine queries.

Does it disclose to users that they're interacting with AI?

We're... in the process of adding enhanced transparency features.

So currently, no?

We're implementing industry-leading transparency measures across all our AI touchpoints.

One more thing — Robert Acheson in Cork says your chatbot gave him incorrect fees and NovaTech wouldn't honour them. Are you aware?

I'm not aware of specific customer cases, but NovaTech takes all customer feedback seriously.

The Central Bank called at 7:15. They've seen the FT article. They want our AI governance framework in 48 hours. Elena — what documentation do we actually have?

We have a ChatGPT subscription and 127 unpublished blog post drafts with no version history.

Then we have 48 hours to build what we should have built 11 months ago. Under regulatory scrutiny, not ahead of it.