A data-security startup just raised $72 million to stop AI leaks. Here's what's already in your bid room's chat history

A data-loss-prevention startup called MIND raised $72 million this week specifically to stop company data from leaking into AI tools. Its own research says 65% of enterprises don't trust the controls they have over that risk today. If your estimators, PMs, or precon team have ever pasted a spec section, a sub's quote, or a set of owner drawings into ChatGPT to summarize it, you're the exposure this raise is about — even though nobody pitched you the product.

What did MIND actually raise, and why does it matter outside IT?

MIND announced a $72 million Series B on September 17, led by Crosspoint Capital Partners with existing backers YL Ventures and Paladin Capital Group, bringing its total funding to $112 million. The company's software inventories sensitive files across SaaS apps, endpoints, and email, classifies them, and blocks or flags risky actions — including sensitive text pasted into generative AI tools or handed to AI agents — before it leaves the company. In its own funding materials, MIND cites survey figures showing 90% of enterprises have deployed generative AI tools, more than two-thirds are using AI agents, and 65% lack confidence in the data controls they have over any of it.

That's not a construction story on its face — MIND sells to large enterprises with security teams. But the underlying problem it's describing doesn't stop at companies big enough to buy DLP software. It's a description of what's already happening at a 60-person GC with no IT department and three project managers who each opened a personal ChatGPT account last year.

What's actually leaking on a construction job?

A 2024 academic review of ChatGPT use on construction projects named the categories most likely to end up pasted into a public AI tool: architectural and engineering drawings, bid documents, cost estimates and profitability calculations, and subcontractor and supplier agreements. None of that is hypothetical — it's the normal working set of an estimator building a bid, a PM answering an RFI, or a precon lead summarizing a spec section. The risk isn't that AI reads the document once. It's what tier of account is doing the reading:

That table is the whole fix for most firms — but it only works if someone has actually standardized which tier the company runs on and told people not to use their personal account for project work.

Should a mid-size GC or sub act now?

Yes, and it costs nothing to start. A firm doesn't need MIND's product — that's built for organizations already juggling AI risk across dozens of SaaS tools. A GC or sub with no security team can close most of the gap with three moves:

1. Put the company on a business-tier AI account (Team, Enterprise, or a gated API tool) instead of letting staff use personal logins.
2. Name the documents that can't go into any AI tool in one paragraph of the employee handbook — bid pricing, owner-furnished drawings, executed subcontracts, anything under an NDA.
3. Tell your bonding company or insurer where you landed, the same way you'd flag a new subcontractor prequalification process. A confidentiality breach through an AI tool is still a confidentiality breach.

MIND's raise is a signal about the size of this gap, not a product every contractor needs. The gap itself — sensitive project data going into whatever AI tool is fastest, with nobody checking which tier it's running on — is already sitting in your bid room.

We covered the federal side of this same AI-vendor-scrutiny trend in the FBI's advisory naming six Chinese AI firms as security risks — that piece is about who built the model underneath a vendor's tool; this one is about what your own team is typing into it.

Forward this to whoever wrote your firm's last confidentiality clause.

Friday one chart. Every week, one piece of data that should change a decision on your project. Subscribe at constructionaibrief.com.