The Delivery Is Evidence Too

2026-10-12 · Philip Choo · AI9OS

Everything up to the send is careful.

Collection is fingerprinted as it happens. The pack is audited, its citations checked, its counts reconciled, its pages looked at, and every file sealed so that any later change would be caught.

Then it is attached to an email, and the chain of custody simply stops.

Afterwards, the only record that a particular bundle reached a particular solicitor on a particular day lives in the sender's own mail client. That is the practice's uncorroborated account of its own conduct, held in a system it controls and could edit, in a form that proves nothing about which version of the attachment went.

Why this matters at exactly the wrong moment

Most of the time nobody asks. The pack went, it was received, the matter moved on.

It matters on the day somebody does ask, and that day tends to be expensive. Disclosure and privilege arguments often turn on when material was first provided, and in what form. "When did your client first see this exhibit?" is a question with a date in the answer. A recipient who says they never received a particular warning, or received a different version of a schedule, can be answered with a verifiable record, or not at all.

And a practice that can prove the integrity of everything up to the envelope, and nothing about the envelope itself, has built its assurance one step short of where it is tested.

It looks complete, which is the problem

The gap is invisible in the usual way for this family of failures. Nothing is missing. Nothing errors. The pack is right there in its folder and the sent email is right there in the outbox. What is missing is the link between them, recorded by anything other than somebody's say-so.

Treat the send as an evidential event

The fix is to register the delivery the same way collection is registered: as an event with its own record.

One record per send, carrying:

• which files were sent, and the fingerprint of each;
• which archive they went in;
• to whom, at which address, and when;
• what assurance those exact files had, meaning which checks they passed and when.

Four rules keep that record honest.

1. Verify the pack before recording the send, and refuse to record if it fails. A delivery record over files that are not the files that passed the checks is worse than no record at all. It looks like assurance while attesting to something untrue. This is where sealing the pack at build time pays off: the send record can only be written over a pack whose seal still holds.

2. One event, one record. Sending the same files to the same address twice by mistake is still one delivery, and should produce one record. The same pack to a different address is a different event and gets its own.

3. Say what the record does not prove. It proves what was sent, to whom, and when. It says nothing about whether the contents are true, since grading does not change because something was delivered, and it says nothing about receipt or reading. Stating those limits in the record itself stops it being used to claim more than it can.

4. Keep a plain copy next to the pack as well. A record that needs a specialist tool to read is not much use to a colleague under time pressure. A simple, readable version alongside the pack is.

A mistake worth remembering

The first working version of this wrote its delivery record into the pack folder itself. That changed the contents of the pack, which made the pack fail its own seal check, which then refused to record the second recipient.

A record about a thing must not be stored inside the thing. It was caught in testing. On a live matter it would have surfaced on the second send, at exactly the moment when a clean record mattered.

The same idea, one step further on

Chain of custody is normally discussed as something that applies to evidence coming in. It is set out in more detail in chain of custody in an era of synthetic media. The point here is that it has a second end. What leaves the practice is also evidence, of what the practice said and when, and it deserves the same discipline as what arrived.

If you are the one receiving the work

"If I later needed to show exactly what you sent me and when, what would you produce?"

"The email" is an answer. It is the practice's own record of its own action. A record fingerprinted at the time of sending, tied to the checks the files had passed, is a different kind of answer, and it is worth knowing which one you would get before you need it.