OpenAI Launches Astra for Law With Legal Search and Plugins

OpenAI is packaging GPT‑6 Astra into a specialized legal platform rather than asking law firms to adapt a general chatbot. Astra for Law combines model reasoning, a legal search index, confidentiality controls and partner plugins, signalling that the next contest in professional AI will be over integrated workflows and verifiable sources—not raw model access alone.

A Legal Configuration Improves Research Performance

According to OpenAI’s announcement, Astra for Law pairs GPT‑6 Astra with instructions for legal analysis and writing plus a search index covering US case law, statutes, regulations, court rules and administrative decisions. The corpus spans more than 230 million URLs and incorporates CourtListener data covering more than 99.9% of published US precedential case law.

OpenAI tested the configuration on 200 private questions from Vals AI’s Legal Research Bench. At the highest reasoning effort, Astra for Law passed the benchmark’s overall correctness check on 54.0% of questions, versus 38.7% for GPT‑6 Astra using web search alone—a 40% relative improvement. It also found 24% more reference cases and retrieved up to 54% more relevant passages in an audited set.

The Product Is a Foundation Rather Than a Closed App

OpenAI describes Astra for Law as infrastructure that firms and legal-technology companies can build around their own expertise. API customers Harvey and Legora can incorporate its capabilities into their products. This avoids forcing established legal workflows into a single OpenAI interface while extending the company’s influence deeper into the sector’s software stack.

The approach reflects the enterprise architecture discussed in Business 2.0’s analysis of AI-native company workflows. The model supplies general reasoning, while legal platforms contribute matter context, permissions, licensed information and review processes. It also creates channel tension: OpenAI can support specialist vendors while building capabilities that overlap with parts of their products.

Plugins Bring Astra Into Existing Firm Systems

OpenAI is launching 26 partner-built legal plugins. The company highlights connections to platforms such as Relativity and Clio, alongside document, timekeeping, knowledge and matter-management tools. These integrations could let lawyers research an issue, draft a brief and save work into the correct matter without repeatedly moving content between systems.

That convenience makes governance more important. Firms need controls over which matters, documents and sources Astra can access, plus records showing what supported an answer. OpenAI says it is expanding privacy and governance controls for confidential client work. The same implementation challenge appears in Microsoft’s guidance for enterprise agents and Google Cloud’s legal AI strategy.

Benchmark Gains Do Not Remove Professional Responsibility

A 54% correctness rate is a meaningful improvement over the comparison system, but it also demonstrates why lawyer review remains mandatory. Legal research can fail through a missing controlling authority, an outdated rule, incorrect jurisdiction or misleading quotation even when the answer reads convincingly. OpenAI’s own wording emphasizes authorities that a lawyer can examine rather than an autonomous legal conclusion.

The American Bar Association’s guidance on generative AI places responsibility on lawyers for competence, confidentiality, supervision, candour and reasonable fees. Firms therefore need independent evaluation by practice area and jurisdiction. Business 2.0’s coverage of Vals AI’s real-world benchmarks reinforces why private testing should supplement vendor-reported averages.

OpenAI Is Competing for the Legal Intelligence Layer

OpenAI’s wider legal offering now extends from general productivity into source-backed research and connected firm operations. The strategic value lies in becoming the intelligence layer beneath multiple legal products, not merely selling ChatGPT seats. Search quality, partner economics and confidentiality controls will determine whether firms accept that role.

Astra for Law gives OpenAI a stronger enterprise proposition, but the product enters a market built on professional accountability and highly sensitive information. Adoption will depend on whether firms can constrain the system, verify every important authority and preserve client confidentiality. Those requirements echo the governance issues examined in AI provenance and accountability standards. In law, better model performance is valuable only when the surrounding evidence and controls are dependable.