Who Pays When AI Agents Make Bad Decisions

The Register examines the legal vacuum around AI agent liability as Oracle, Microsoft, SAP, Salesforce, and Workday deploy agents that autonomously handle HR screening, regulatory filings, and supply chain decisions. None of these vendors would comment on whether they accept liability for agent errors. The UK's Financial Reporting Council put it bluntly: you can't blame it on the box. Gartner's Lydia Clougherty Jones projects AI-related remediation costs exceeding $10 billion by mid-2026. The contract problem is structural - vendors can't warranty non-deterministic behavior, and existing legal frameworks weren't built for systems where fault splits between model bias, algorithm design, and user prompts. Linklaters partner Georgina Kon captures the bind: you can't have people deploying autonomous agents and not being liable, but the law hasn't caught up to say exactly who is.