Open model inference is no-retention by default... Responses API up to 30 days, batch files until customer deletion
Original headline오픈 모델 추론은 기본 무보관…Responses API는 최대 30일, 배치 파일은 고객 삭제 때까지
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
Fireworks AI CEO Lin Qiao clarified on the 14th that the company's zero-retention promise for open-model inference covers only standard inference, where inputs and outputs exist solely in volatile memory and are never written to persistent storage, while its Responses API retains data for up to 30 days and batch-job datasets uploaded to cloud storage persist until the customer deletes them. The distinction drew scrutiny after Notion AI chat histories surfaced in a Fireworks security incident dataset, prompting debate over the scope of the no-retention commitment. Qiao likened uploading files to cloud storage to entering numbers into a calculator, framing batch data as customer-managed rather than transient. The disclosure matters because enterprise buyers increasingly weigh data-retention terms when selecting inference providers for sensitive workloads. What to watch next is whether Fireworks extends zero-retention guarantees to its Responses API and batch products, and how rivals position their own retention policies.
Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.
Named in this story
Fireworks AI, Lin Qiao, Notion AI