Exports, list freshness, and a screening coverage guarantee — what shipped this week
A busy week: v1.37 to v1.43, six releases, and two of them fixing things we’d rather you hear about from us than find yourselves. Here’s what’s different when you log in.
You can export things now
Until this week, the only thing you could take out of ARGOS was a chat answer as Markdown. Now:
- Case reports download as PDF or Markdown from the case detail page — entity, individuals, red flags, the full report, sources, and the status history.
- The audit trail downloads as CSV, respecting whatever event-type filter you have applied. Every cell is neutralised against spreadsheet formula injection, so a name that happens to start with
=or+opens as text, not as a formula. - A screening result exports as a PDF evidence pack — matches, the AI adjudication, adverse-media leads — from the button next to “Copy JSON”.
We chose the PDF library specifically for security: it draws the document directly rather than rendering HTML, so there’s nothing in an exported report that could be tricked into fetching an outside URL. The trade-off is plainer typesetting. We’ll take that.
Our own security and SLA pages used to say self-service export was “on our roadmap.” They don’t any more.
Source Lists shows when each list last refreshed
Each sanctions list on Source Lists now carries an “Updated” column — how long ago that list actually last synced — and turns amber if it’s been more than 48 hours. If the nightly sync degrades two runs in a row, every super-admin gets an email.
We added this because of something we found ourselves. Canada changed the format of its SEMA sanctions feed in late August, our parser stopped recognising the records, and for eleven nightly runs the sync loaded zero Canadian entries while reporting success — because a list that parses to nothing looked identical to a list we’d deliberately switched off. Screening kept working from the last good snapshot, so no hits were lost, but the list was stale and nothing told us. Now a zero-record sync fails loudly, and you can see the freshness yourself.
No sanctions list can be dropped by the result cap
A screen used to return its top five matches by score. For a heavily sanctioned name — someone on six or seven lists at once, all scoring 100 — that meant whichever lists sorted last simply didn’t appear. We found it on a well-known name that came back with four sanctions lists and neither the OFAC nor the UK FCDO entry.
Two changes: the cap is now 25, and the screening engine keeps the best match from every list before the cap applies at all. A name on eight lists shows eight lists. The cap still stops eight aliases from the same list crowding the page.
Smaller things
- The audit trail now records which AI model produced each chat answer and investigation report — shown as a chip next to the event.
- Pages load with placeholders instead of briefly showing “no events logged yet” while data is still on its way.
- Team invites are emailed to the invitee as well as shown as a link to copy.
- The Models page’s Enable/Disable buttons work again, and signing in lands you on your dashboard rather than bouncing off an admin page.
If something here doesn’t behave the way this describes, tell us — that’s how the two fixes above got found.