Agents of Failure

Henry Goodstone builds the case against deploying AI agents faster than the governance meant to contain them, using court records, engineering post mortems, security reports and parliamentary inquiries. From the founder who watched an agent delete his production database to 26,000 Dutch families wrongly accused of fraud, the failures are not edge cases but the predictable cost of shipping a technology that is not ready. The book climbs from one person alone with a consumer app up to government scale, and asks the question the industry avoids: who pays when the agent fails, and who is answerable for it. It is not an argument against AI. It is an argument for honesty.