Verified facts as of September 3
- OpenAI release notes name GPT-6 Astra and say rollout is reaching a limited set of organizations; the same note says it is not yet generally available and broader availability is planned over coming days.
- OpenAI’s Legora customer story reports 41 financial-statement documents reviewed in minutes, a 40% improvement against Legora’s benchmark and all four planted errors identified; these are vendor-reported workflow results, not an independent legal-outcome guarantee.
- OpenAI’s Playco customer story reports 50% fewer manual fixes across three prototypes; it is supporting rollout evidence from a non-legal workflow, not proof of law-firm performance.
- OpenAI’s Path to Astra says Astra meets the Critical cybersecurity capability threshold, while the displayed results reflect Daybreak Blue access rather than default production.
- An attributed product-lead post says release is starting and expanding over days; it does not establish access for this account, organization or region.
Fields that remain Unknown
| Field | Status | Release evidence needed |
|---|---|---|
| General availability date | Not yet generally available; exact date Unknown | OpenAI product announcement |
| Account, organization and region access | Unknown | Account surface and availability evidence |
| API model ID and price | Unknown | OpenAI API documentation and pricing |
| ChatGPT and Codex plan access | Unknown | OpenAI availability matrix |
| Business-data retention for Astra | Configuration-specific | Operative product terms and account controls |
| Legal-work accuracy | Unknown | Disclosed legal workflow test |
Prepare six questions before launch
- Which Astra surface and model configuration will the firm use?
- Which safety checks can pause or stop legitimate work?
- Which tools, code repositories and external systems can the model reach?
- Which business-data retention and human-review terms apply to that surface?
- Who authorizes an action and who reviews the result?
- How will the team preserve logs, revoke access and respond to a model or security incident?
Keep legal-work claims separate
OpenAI’s September 1 post evaluates cybersecurity and safeguards. It does not establish legal research accuracy, contract-drafting quality, privilege treatment or compliance with a jurisdiction’s professional rules. A future legal-work pilot needs its own sources, rubric and reviewer.
Monitoring rule
Update this page when OpenAI changes rollout scope, publishes a product announcement, system card, pricing or access documentation. Treat the release notes as the primary status source and the social post as an attributed supporting signal. Do not turn limited rollout into universal access. Preserve this dated status record so readers can distinguish it from a later general-availability record.
FAQ
Is OpenAI Astra available now?
OpenAI’s release notes say GPT-6 Astra is rolling out to a limited set of organizations and is not yet generally available. Account-specific access remains Unknown until verified.
Is Astra GPT-6?
OpenAI’s September 3 release notes use the name GPT-6 Astra. The account route, API model ID and pricing remain Unknown without direct documentation.
Can law firms prepare during the rollout?
Yes. Define approved data, tools, human authorization, review, retention questions and incident handling, then wait for route-specific launch evidence before testing matter data.
Sources checked
- ChatGPT release notes · GPT-6 Astra rollout, checked 2026-09-03.
- Legora financial-statement review with GPT-6 Astra, checked 2026-09-03.
- Playco game prototyping with GPT-6 Astra, checked 2026-09-03.
- Path to Astra, checked 2026-09-03.
- Astra rollout signal, checked 2026-09-03.
- Business data privacy, security and compliance, checked 2026-09-03.
Operational information, not legal advice. Verify current terms, account configuration and applicable professional duties before use.