AIToday
Large Language ModelsAI Safety & AlignmentSiliconANGLE AIPublished: Sep 18, 2026, 06:00 JST

AI agents erase audit trail, Vast Space audit chief warns

AI agents erase audit trail, Vast Space audit chief warns

3 Key Points

  1. What happened

    Speaking at Workiva's Amplify conference, Vast Space chief audit executive Josh Robinson said AI agents remove the email threads, Slack messages and tick marks that once let auditors see human judgment, so the risk sits in the unknown.

  2. Why it matters

    The loss of visible evidence lands hardest in regulated industries where AI governance is now a board-level question, and ownership of signed audit work still sits with the person who signs it, whichever tool produced it.

  3. What to watch

    Robinson argues the fix is old discipline, not new tooling: applying tests of completeness and accuracy step by step across a multistep agent workflow. Watch whether that holds as agents move faster than any reviewer can follow.

WHO IT HITSChief audit executives and internal audit teams in regulated industries are affected, since they must sign off on work whose reasoning is increasingly produced by software rather than recorded by people. Enterprise finance and HR teams running data across NetSuite, HR systems and data lakes may also face pressure to connect those silos first.

Not sure about something? Ask the AI

Summaries like this, in your inbox every morning.

Context & Analysis

The shift Robinson describes is not a new tooling problem so much as a record-keeping one. Financial and operational data still sits across NetSuite, human resources platforms and data lakes, but the reasoning that once surrounded that data — the handwritten notes, tick marks, email chains and Slack messages — is being produced, if at all, by software moving faster than any reviewer can follow. That is why the issue surfaced as a board-level question at Workiva Inc.'s Amplify conference rather than as a compliance checkbox.

Robinson's proposed remedy is deliberately old-fashioned. Internal audit has spent years pressing completeness and accuracy, and he argues those same tests can be applied step by step across a multistep agent workflow. The harder problem is accountability: at Vast Space LLC, which builds commercial space stations and carries astronaut safety risk alongside regulatory exposure, control gaps have consequences well beyond a restatement, and ownership stays with the person who signs the work.

Workiva's pitch fits that gap. Its specialized AI agents and Knowledge intelligence layer aim to provide one connective layer spanning NetSuite, HR systems, data lakes and whichever model a company has chosen — Robinson, a customer since 2017, describes asking a CFO to look at a single layer connecting those systems. Whether that resolves the unknown he worries about is likely to hinge on how completely those silos are actually connected, and on whether auditors can still demonstrate completeness and accuracy over work they did not watch being done.

FAQ
Why do AI agents make auditing harder?
Josh Robinson of Vast Space said that with humans auditors could see handwritten notes, tick marks, email chains or Slack messages showing judgment. That evidence has all been removed, so the risk is in the unknown.
What does Robinson say audit teams should do about it?
He says the answer is not new tooling so much as old discipline. Demonstrating completeness and accuracy step by step across a multistep agent workflow generally lets auditors reach conclusions about assurance of datasets or audit tests.
Who remains accountable for AI-produced audit work?
Robinson said ownership stays with the person signing the work, regardless of which tool produced it. At Vast Space, which builds commercial space stations, control gaps carry consequences beyond a restatement.
SiliconANGLE AIRead Original Article

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Mastercard Joins Visa in Letting AI Bots ShopTop Companies AI · 1h ago
  • Mastercard AI payment tool lets bots shop without approvalTop Companies AI · 1h ago
  • JPMorgan caps Claude at $2,000 a month for some engineersTop Companies AI · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleTopview Optronics sees AI supply-chain shifts lifting orders through end of 2026