product updateOpenAI

OpenAI Reaffirms Zero Data Retention for API Customers, Previews New Private Safety Processing

TL;DR

OpenAI has reaffirmed its Zero Data Retention (ZDR) policy for eligible API customers using frontier models and previewed a new capability called Private Safety Processing. The company says the new approach aims to preserve safety monitoring capabilities without requiring data storage.

2 min read
0

OpenAI has reaffirmed its commitment to Zero Data Retention (ZDR) for eligible API customers using its frontier models, and previewed a new capability called Private Safety Processing designed to preserve safety oversight without storing customer data.

Zero Data Retention is an existing API option that allows qualifying enterprise customers to send requests to OpenAI's models without the company retaining inputs or outputs after processing completes. OpenAI has offered ZDR to select customers for specific use cases in the past; this announcement reaffirms and expands the company's commitment to the policy for its current frontier model lineup.

According to OpenAI, eligibility for ZDR depends on use case and customer agreement terms — the company has not disclosed a full list of qualifying criteria or which specific models and endpoints are covered. Pricing for ZDR-eligible API access has not been separately disclosed; it is unclear whether ZDR carries a premium over standard API pricing.

The more notable element of the announcement is Private Safety Processing, which OpenAI describes as a preview capability. The company claims this system will allow it to run safety classifiers and abuse-detection checks on API traffic without retaining the underlying customer data used to perform those checks. Details on how this is architected — whether through on-device processing, ephemeral compute enclaves, or cryptographic techniques — were not specified in OpenAI's announcement. No technical paper, benchmark, or third-party audit was cited to substantiate the privacy guarantees of the new system.

OpenAI has not published a rollout timeline for Private Safety Processing, nor specified which customers or tiers will get access first. The company frames the effort as resolving a tension between two demands from enterprise customers: strong safety monitoring to prevent misuse, and strict data privacy guarantees that prevent OpenAI from retaining sensitive inputs.

What this means

This is a policy and trust announcement, not a new model or technical breakthrough. OpenAI has faced pressure from enterprise and government customers — particularly in regulated sectors like healthcare, finance, and legal — who require contractual guarantees that their data won't be stored, used for training, or accessible to OpenAI staff. ZDR has existed in some form since at least 2023, and this reaffirmation is likely aimed at reassuring customers amid growing competition from Anthropic, Google, and open-weight providers who have made similar or stronger data-handling commitments.

The more interesting signal is Private Safety Processing. If OpenAI can genuinely run abuse detection without retaining data, it addresses a real architectural problem: safety monitoring traditionally requires inspecting content, which conflicts with strict no-retention promises. But until OpenAI publishes technical details — what's actually stored, for how long, and how classifiers operate on data they don't retain — this remains a claim rather than a verified capability. Enterprise buyers evaluating this for compliance purposes (HIPAA, GDPR, SOC 2) will want independent audits before treating it as a settled guarantee, not just a marketing preview.

Related Articles

research

OpenAI Reports Up to 14% PPA Gain on AMD Flagship GPU IP From Model-Rewritten RTL

OpenAI reported on November 18, 2026 that its models achieved up to a 14% improvement in power, performance and area (PPA) on AMD flagship GPU intellectual property, measured against heavily optimized baselines. Some of the circuit rewrites were included in a design sent for manufacturing.

research

OpenAI models optimized circuits used in Jalapeño silicon and improved RISC-V core designs

OpenAI reports that its models optimized circuits used in Jalapeño silicon and improved RISC-V core designs. The company presents the work as progress in AI-assisted chip design with GPT-6 Astra.

product update

Microsoft makes Word, Excel and PowerPoint in Copilot generally available, calls Copilot "the new OS for work"

Microsoft CEO Satya Nadella announced Tuesday at Ignite the general availability of Word, Excel and PowerPoint inside Copilot, which he called "the new OS for work." The company also introduced Copilot Autopilot, Microsoft IQ and expanded Agent 365 controls, with pricing for Autopilot still unannounced.

product update

Replit launches Missions: agent swarms work toward goals within budgets and approval limits you set

Replit has launched Missions, a product where users define an outcome and Replit coordinates a swarm of agents to plan, build, test and iterate within a budget and boundaries the user sets. Replit says it used Missions to build an application monitoring platform, DataCat, in 2 days within a $3,000 budget.

Comments

Loading...