Estimated reading time: 5 minutes · Last updated:
OpenAI has launched Astra for Law, a version of its generative-AI tools tailored for big-firm legal work and aimed at the 200 largest US law firms known as AmLaw 200. Astra for Law combines case-searching tools, plugins for document and research providers, and model configuration intended to reduce hallucinations and produce longer, more practice-oriented answers. OpenAI says the product runs on its most capable model and that legal teams, third-party legal-tech vendors and document platforms can integrate it. The move is part of a broader industry push by OpenAI into sector-specific products, and it was announced as first reported by Business Insider.
You win by focusing on the customer and not the competition.
Jason Boehmig, who leads legal product at OpenAI
Key takeaways
- Product and target: OpenAI launched Astra for Law to target the AmLaw 200, the 200 largest American law firms.
- Technical base: OpenAI says Astra for Law runs on the company's most advanced model and includes model tweaks tuned to legal work.
- Longer answers: OpenAI said Astra for Law returns answers roughly twice the length of standard Astra outputs, a length chosen to fit legal workflows.
- Third-party integrations: Legal tech vendors including Harvey, Legora, Thomson Reuters, iManage and DeepJudge will supply ChatGPT plugins or add Astra for Law to their products.
Table of contents
Why OpenAI is packaging a law-specific product
OpenAI has shifted from a general-purpose assistant to industry-specific offerings, and Astra for Law is its legal-market play. The company is pursuing customers by embedding tools and configuration changes that its product team says make the output more relevant to legal practitioners. Jason Boehmig, who leads legal product at OpenAI, framed the strategy around customer focus rather than head-to-head competition, and the product is pitched at the workflows used by firms in the AmLaw 200.
The legal tweaks emphasise factual grounding in statutes and case law and longer-form answers. Sherwin Wu, a member of technical staff at OpenAI, said the model adjustments ground the system in “actual case law,” and that this reduced hallucinations and improved reasoning. That technical aim matters because law firms treat mistaken or unsupported assertions as an immediate liability and demand explainability tied to authorities.
OpenAI also sees a commercial logic: law firms are high-value, repeat customers with established document and research systems that lend themselves to paid integrations and subscription arrangements. The company paired the technical changes with integration points—plugins and connectors—so existing legal-software vendors can embed the model into platforms lawyers already use.
How Astra for Law works inside legal workflows
Astra for Law bundles tools intended to map into common legal tasks: faster case search, longer analytic outputs, and connectors to practice systems. The initial launch includes ChatGPT plugins from Harvey and Legora and plugins or integrations from Thomson Reuters, iManage and DeepJudge, allowing firms to surface model answers alongside their document stores and research feeds.
One feature OpenAI plans to bring from its financial-services rollout is a citations feature that links assertions to sources; the legal product team said they will add a citations feature to Astra for Law soon. Integrations let vendors call the model from inside matter-management, document-review and research workflows so the model's outputs can be checked against a firm's own databases and vendor libraries.
That integration strategy lets vendors add Astra for Law as an embedded capability without replacing existing platforms. Legal-tech companies can add the tool to their products and offer model-driven drafting, summary, or research assistance while keeping custody of firm content and review trails.
What Astra for Law means for legal-tech vendors and law firms
Vendors such as Harvey and Legora can embed Astra for Law into their products to offer model-augmented drafting and research; Thomson Reuters, iManage and DeepJudge are listed as plugin partners. For vendors, that option shifts the decision from building proprietary models to integrating a specialist model and focusing on workflow and compliance controls.
For AmLaw 200 firms, the promise is faster research and longer, more detailed answers that map to legal standards. OpenAI's technical staff described the product as tuned to reduce hallucinations by anchoring outputs in authority material, which is a direct response to a core client concern. Firms will still need review processes and integration with their document management, but the vendor route can accelerate deployments.
The launch also raises commercial questions for incumbents: how to price embedded AI, how to certify model outputs for regulatory and malpractice risk, and whether to partner or compete with model providers. OpenAI positions Astra for Law as a platform for legal work that third-party suppliers can add to their offerings rather than a single monolithic replacement for existing tools.
| Vendor | Role at launch | What the material says |
|---|---|---|
| Harvey | Legal-tech vendor | Can add Astra for Law to its products; supplies a ChatGPT plugin |
| Legora | Legal-tech vendor | Can add Astra for Law to its products; supplies a ChatGPT plugin |
| Thomson Reuters | Research and content provider | Provides a ChatGPT plugin at launch |
| iManage | Document-management provider | Provides a ChatGPT plugin at launch |
| DeepJudge | Legal-tech vendor | Provides a ChatGPT plugin at launch |
Upside and downside for adoption
The case for
- Firms can speed routine research and drafting by integrating Astra for Law into existing document and practice-management systems.
- Vendor partnerships let legal-tech companies add model capabilities quickly without building from scratch, reducing time-to-market for AI features.
The case against
- Firms will still require human review and validation because legal standards and malpractice exposure do not disappear with model tuning.
- Commercial and compliance questions—pricing, auditability of outputs and liability allocation—could slow enterprise adoption among the AmLaw 200.
What to be careful about
- Model errors or unsupported assertions will remain a production risk for law firms even if hallucinations fall, and firms must retain review workflows.
- Unspecified pricing, deployment and certification terms could delay procurement decisions by AmLaw 200 firms and their vendor partners.
- Vendor integrations may create variation in accuracy and controls depending on how each partner implements Astra for Law inside its product.
The bottom line
Astra for Law is OpenAI's focused effort to bring generative models into major law-firm workflows by pairing model tuning with vendor integrations. The product targets the AmLaw 200 and promises longer, more authority‑anchored answers plus plugins from established legal-tech and research providers. That approach reduces the engineering burden on vendors but leaves open commercial and compliance questions—pricing, auditability and malpractice risk—that firms and vendors must resolve. Adoption will hinge on whether OpenAI and its partners can prove the model's reliability in live matters and on how quickly legal teams accept embedded AI into their review and sign-off routines.
What to watch
- watch whether AmLaw 200 firms announce pilot programs for Astra for Law; no date has been set.
- watch for the rollout of the Astra for Law citations feature across vendor integrations; OpenAI said the feature will arrive soon but provided no date.
- watch which legal-tech vendors list Astra for Law as generally available in their products; no date has been set.
Frequently asked questions
What is Astra for Law?
Astra for Law is OpenAI's legal-focused offering that combines model adjustments, case search and ChatGPT plugins to serve law-firm workflows; OpenAI said it runs on the company's most advanced model.
Which firms is OpenAI targeting with Astra for Law?
OpenAI said it is targeting the AmLaw 200, the 200 largest American law firms, and positioned the product for integrations into existing vendor platforms.
How does Astra for Law differ from recent legal releases by Anthropic?
Both companies released legal-focused tools with integrations; OpenAI says its product includes model tweaks to ground outputs in case law and plans to add a citations feature drawn from its financial-services launch.
Related reading
