OpenAI unveils Private Safety Processing, a privacy-centered safety system designed to detect potential misuse of AI models without accessing or retaining customers’ underlying data.

ZDR (Zero Data Retention) aims to keep user input on-device, with a future option to encrypt and store input on OpenAI infrastructure using user-only keys.

OpenAI commits to Zero Data Retention for frontier models, ensuring prompts and responses aren’t retained or reviewed by OpenAI staff unless customers opt in.

Relativity positions claiR as part of a broader push to meet legal professionals where they work, accelerating early case assessment and decision-making.

Adjacent market news includes Abridge expanding clinical intelligence, Costco entering Medicare with SCAN Group, and a high-profile covid-era research guilty plea, framing the healthcare AI and policy context.

Big-picture takeaways: agentic AI adoption, cost-aware routing, and governance controls are the key drivers of enterprise AI success in 2026.

Organizations are increasingly deliberate about where to deploy different models based on cost, reliability, and reasoning capabilities.

Updates extend the agent beyond note generation into revenue cycle management, with over 400,000 hours saved across U.S. health organizations since launch; newer features drive growth.

Oracle Health’s updates unfold in the wider AI arms race among EHR vendors, with Epic’s Ergo Visit launching concurrently as AI tools expand in the ecosystem.

Risks include execution challenges in complex healthcare deployments, accuracy, compliance, and clinician trust, which will shape adoption and monetization of AI features.

Relativity plans to share more details at RelFest Chicago (Sept. 29–Oct. 1) and signals commoditized pricing on general availability in 2027, with sign-ups for updates on claiR.

Relativity expects claiR to be included in RelativityOne at no extra cost when generally available in 2027; early testing involves top law firms.