Nobody Checked Deloitte’s Report. One Academic Did. It Cost Them A$97,587.

Source: Nate’s Newsletter Date: 2026-08-05 URL: https://natesnewsletter.substack.com/p/ai-slop-cost

Summary

Nate’s Newsletter uses a Deloitte report — an academic caught unreviewed AI-generated errors, and it cost the firm A$97,587 — as the anchor for a piece on who actually pays for unvetted AI output. The author’s frame: “the sender got the speed, you got the bill.” Rather than arguing against AI use, the piece pushes for treating authorship as deliberate — AI as a tool inside a verification process, not a replacement for one.

Implications

Feeds the cost/economy of AI thread from an unusual angle: the “abundant intelligence” story (cheaper tokens, GPT-5.6 price cuts) has a cost accounting that only tracks the seller’s side. This piece prices the buyer’s side — verification labor and reputational damage — as a real, sometimes large, line item that doesn’t show up in per-token pricing. It’s also adjacent to verifiability: the Deloitte incident is exactly the failure mode that formal-verification and disclosure norms (Lean-certified proofs, eval-escape self-disclosure) are trying to price out of the system before a client, not an academic, has to catch it.