IRON is the research pipeline built for a gap that has existed since AI became a business-critical layer in every acquisition target. We built the instrument because nobody else did.
Traditional diligence rests on the documents the target hands you. Privacy policies in the data room. Compliance questionnaires answered by counsel. Vendor lists vetted by the CFO.
Those documents describe what the target has chosen to say about itself. They don't describe what the target actually runs. For most of an operating company, those two pictures roughly match. For the AI layer, they diverge.
The AI stack moves faster than disclosure. Engineering teams add models, swap vendors, and ship features between board meetings. Privacy policies get updated on a quarterly review cadence. Sub-processor lists get updated when legal remembers. The result: a structural blind spot in standard M&A diligence sized to whatever the AI footprint of the target is.
Existing AI diligence tools (Harvey, Kira, Luminance) analyze the documents you receive. They make document review faster. They do not surface what the target chose not to put in a document. That gap is the reason IRON exists.
IRON reads the public signal: the GitHub repos that confirm dependencies, the job postings that confirm production systems, the privacy policy clauses that confirm data routing, the open sub-processor lists. Signals the target has already published to the world. Nothing privileged. Nothing the target needs to cooperate on.
Every quarter without a diligence standard is a quarter of compounding exposure. These are the specific dates deal teams are mispricing right now.
A target operating in New York, Colorado, and Illinois simultaneously sits under three distinct AI compliance regimes with different audit, notice, and documentation requirements. None of them appear in a standard data room. IRON surfaces all three before you close.
See how IRON maps this →IRON is the agent and pipeline name. It is not what the customer buys. Customers buy a written IRON report from Velxa: 20 to 40 pages, counsel-ready PDF, formatted for the data room. IRON does the research silently. The report is the artifact.
IRON's confidence comes from a narrow definition of what it does. It reads public signals, maps every AI system found to the regulations that apply, and writes the findings into a report that counsel can validate against the cited evidence. It does not produce legal advice. It does not access private systems. It does not read the target's data room.