Zhipu AI says it will introduce a no-data-retention option for its model-as-a-service platform, meaning company and developer prompts, along with model outputs, would not be stored after processing. For businesses weighing whether to send contracts, customer information or source code to an AI system, that is a meaningful change, not merely another model benchmark wearing a tie.
Zhipu AI / Z.ai GLM Watch analysis
What happened
Zhipu AI, known as Z.ai in overseas markets, announced the planned service arrangement in remarks reported by China Daily Hong Kong on 22 September. The company says inputs and outputs will not be retained after a request has been processed.
The report frames the move as part of a wider contest among Chinese AI providers over trust, privacy and enterprise adoption. It also notes that China requires generative AI providers to protect users’ input information and usage records, while officials have warned that prompts can contain personal information and commercial secrets.
Why it matters
Retention rules can decide whether an AI service is allowed near sensitive business material at all. A bank, manufacturer or software company may care less about shaving another fraction from a benchmark and more about whether a prompt survives the request, is reused for training or can later be exposed.
Zhipu’s proposal therefore targets one of the least glamorous but most consequential parts of AI deployment: what happens to the data after the clever answer arrives. The supplied report describes a planned option, not evidence that every account already has it or that all connected services follow the same policy. Privacy, as ever, lives in the settings menu and the footnotes.
Our read
This is a useful enterprise signal because it turns data handling into a product feature rather than leaving it as a promise buried in governance language. But buyers should ask for the operational detail before treating “no retention” as a complete security guarantee.
The practical question is not simply whether Zhipu stores ordinary requests. It is which endpoints, logs, backups, abuse-monitoring systems and legal exceptions are covered, and how a customer can prove the promised setting was active. A clear retention policy can unlock serious use. A vague one merely gives procurement teams another document to circulate.
What to watch
- Whether Zhipu publishes activation rules and an effective date for the option.
- Whether batch processing, file uploads and other persistent workflows are excluded.
- How the company handles abuse investigations, legal requests and operational logs.
- Whether customers receive audit evidence rather than a reassuring sentence on a webpage. Zhipu’s announcement matters because confidential data is becoming one of the main gates on enterprise AI adoption. The next test is whether the controls are precise enough for a security team to sign off, rather than simply pleasant enough for a sales deck.
Discussion spark: Should AI providers be required to make no-retention processing the default for enterprise accounts, or should customers accept the burden of configuring and auditing it themselves?
Sources and evidence
- Zhipu AI ramps up efforts for data privacy – chinadailyhk (22 September 2026, 13:49 UTC)
not affiliated with or endorsed by Zhipu AI, Z.ai or GLM