A Sheaf record

The AI Loss Ledger

A curated, sourced record of what happens when an AI agent or an LLM goes astray and it costs someone real money. Every documented case where AI caused a company, or a person, a material loss: financial, legal, or reputational. Every entry links to primary reporting.

documented incidents
largest single documented loss
years covered

This ledger is curated and human-verified. Figures are as reported by the linked sources and reflect widely-published reporting at the time; some are estimates or contested. Entries describe reported events and outcomes, not legal conclusions. Spotted an error, or a case we're missing? Tell us and we'll update it.

This is the problem Sheaf exists to prevent.

Most of these losses trace to the same root: an AI acted, and nobody was measuring whether its reasoning held together before it did. That measurement is what Sheaf is, the coherence layer for agentic AI.

See it measure your agents