Astra for Law is OpenAI's legal configuration of GPT-6 Astra. It combines the model with a legal search index and instructions for legal analysis and writing. For in-house counsel, the question is how it fits the work your team needs to finish.
GC AI is an enterprise legal AI platform built for in-house counsel, used by 2,100+ legal teams as of September 2026. Its CEO and co-founder, Cecilia Ziniti, is a three-time general counsel. GC AI is participating in the Astra for Law launch.
We’re excited to bring GC AI’s legal intelligence to ChatGPT through our GC AI ChatGPT Plugin, and we’ll continue innovating with the model for our customers. For teams evaluating the launch, GC AI offers an established set of contract review workflows to compare against the OpenAI setup they could deploy.
Is Astra for Law Available to In-House Legal Teams?
Astra for Law is initially available to selected law firms through Trusted Access in ChatGPT and Codex. General direct access for corporate legal departments is not described. API access is coming soon, and any partner-product route should be evaluated separately from direct access to Astra for Law.
What Astra for Law Adds
Astra for Law includes U.S. research across cases, statutes, regulations, court rules, and administrative decisions.
Sullivan & Cromwell's agreement analyzer brings the firm's negotiating playbooks and selected precedents into contract review and proposed redlines. This is custom firm work, with broader availability to be confirmed. OpenAI also announced 26 partner-built plugins and general availability of ChatGPT for Word.
The access routes differ:
Legal model: Initially available to selected law firms through Trusted Access in ChatGPT and Codex; corporate legal departments should not assume direct access
Word and plugins: Confirm the enabled model, permissions, and any separate subscriptions; access to an integration does not establish access to the legal model
API: Astra for Law API access is coming soon, so availability and commercial terms need confirmation before planning a deployment
What the Astra for Law Benchmark Tells You
The Astra for Law evaluation reports 54.0% overall correctness versus 38.7% for base GPT-6 Astra with web search. It used 200 U.S. legal research questions from the private Vals Legal Research Bench validation set, with both systems at their highest reasoning effort.
The result measures the configured research system, not the model alone.
It is neither a production error rate nor a comparison with GC AI. To assess its value for your team, check the research and corrections needed to finish a representative memo or redline.
Confirm Data Controls Before Using Company Documents
For eligible firms, the legal offering includes zero data retention for API usage and an eyes-off human-review posture by default for ChatGPT Enterprise.
Its enterprise privacy commitments separately cover no training on business data by default and configurable Enterprise retention.
GC AI’s security documentation states that model providers cannot train on customer data and that we maintain zero-data-retention agreements with LLM providers wherever feasible. We also document encryption, customer data segregation, and SOC 2 Type II assurance.
For either setup, confirm which services receive documents, who can access them, how long they are retained, and who approves new connections. No training, retention limits, and restrictions on human review are separate controls. Check the terms for the products and connected services your team will use.
How GC AI Supports the Work From Research to Redline
GC AI is excited to participate in the launch of Astra for Law through the GC AI ChatGPT Plugin. Try it with a low-risk legal question or a de-identified document if it is available to your team.
It brings GC AI’s legal intelligence into ChatGPT, so you can work from the legal context already in your GC AI workspace. Verify the result against your approved sources and workflow. We’ll continue innovating with the Astra for Law model to improve legal reasoning, research, and agentic workflows for our customers.
At Columbia Sportswear, marketing needed an event liability waiver in less than a day. The existing form did not cover the risks. GC AI helped identify issues and draft language, and the attorney delivered a new form within hours.
Melissa M. Robertson, Columbia's associate general counsel, describes GC AI as:
“helping Columbia’s legal team extend its capacity without adding headcount.”
Research the Question and Check the Sources
GC AI Research uses web search for straightforward lookups and Research Agents for more complex questions, including work alongside uploaded files. For multi-part questions, several agents can investigate in parallel. Our workflow compares and reconciles their findings before returning cited analysis.
For example, counsel could ask which employee-notice requirements apply to a workplace-monitoring policy in California, New York, and Texas. Review the jurisdiction-specific conclusions, unresolved facts, and supporting sources. More agents alone do not establish a better answer.
https://www.youtube.com/watch?v=ERfAOfAJCJg
U.S. Case Law lets you search opinions, filter by court and state, and open cited cases in the full-text Opinion Reader. Midpage supplies the opinions and treatment flags where available.
Apply Your Company's Positions
Company profiles carry business context and risk preferences into chats and projects. Playbooks apply defined standards and fallback positions across agreements, including in Word. You can start with a prebuilt playbook or create one from templates and negotiated agreements.
Keep preferred positions, acceptable fallbacks, and signed obligations distinct. A concession on one deal should not automatically become the next agreement's default. When a finding depends on specific wording, GC AI’s Exact Quote opens the source document with the quoted language highlighted so counsel can inspect the surrounding provisions.
Finish the Review in Word
GC AI for Word supports issue spotting, drafting, redlining, and summaries of comments and counterparty changes.
Hayley McAllister, senior counsel and head of commercial legal at Jasper, said:
“Once the Word plugin rolled out, I pretty much exclusively started using it for all of my redlining and contract review.”
She also used GC AI to refine comments explaining Jasper's positions and develop sample comments for its commercial playbook.
https://www.youtube.com/watch?v=GbygtIhhapk
Identify the Terms in Force
For work spanning multiple signed documents, GC AI’s Contract Intelligence links amendments, statements of work, and renewals to parent agreements. Its Current Terms view brings together the terms in force across each agreement family, with cited extracted values available for inspection. Confirm that the relevant documents are present and scope Contract Intelligence separately in your proposal.
Evaluate the Tools With One Matter Your Team Knows
Use a vendor renewal to test research and contract review together. Start in an approved environment you can access. If your team gains Astra access, reuse the same materials and substantive instructions. These are proposed tests, not results from a hands-on comparison.
Prepare a de-identified master services agreement (MSA), its executed amendment, and an approved playbook. Choose a legal question your team has already researched, such as the effect of a forum-selection clause under a specified jurisdiction. Refresh counsel's reference answer before testing.
Research: Ask for leading and contrary authorities, pinpoint support, material assumptions, and unresolved questions; check the sources against counsel's reference research
Contract review: Ask which terms are currently in force, where they depart from the playbook, and which decisions require approval; include an amendment that changes a limitation of liability exception
Redline: Carry selected findings into Word and check whether the edits preserve negotiated exceptions, defined terms, and cross-references
Repeatability: Change one approved playbook position and confirm that the next review uses it
Keep the reference answer separate from the materials supplied to the system. Record material omissions, unsupported conclusions, and the lawyer's correction time. Note the model, tools, and provider assistance used. A useful pilot reveals whether the setup can support your work; one successful matter does not establish accuracy across the department.
Budget for the Full Workflow
The Astra for Law launch page and law offering page do not specify a price. Base-model API rates are not a quote for the legal offering.
For in-house teams, total cost depends on document length, research steps, connected services, usage allowances, overages, implementation, and lawyer review time.
Ask for an estimate for an ordinary month and a busier month. Confirm whether the proposed environment bundles usage, passes it through, or combines a subscription with allowances and overages.
Request an estimate for an ordinary month and a busier month using your team's expected workload:
Included: Recurring fees, features, usage allowances, and overages
Variable: Connected services, retries, and implementation support
Controls: Cost visibility, alerts, caps, and what happens when a cap is reached
Internal: Lawyer review time and maintenance of approved positions
Our pricing lists Individual at $500 per month and Team pricing on request. U.S. Case Law is an Individual add-on and a Team inclusion. A subscription provides a known starting cost; confirm the included scope, any usage limits or overages, and Contract Intelligence in your proposal.
Start With GC AI for Recurring In-House Work
We recommend starting with GC AI for recurring in-house research, contract review, and drafting. Evaluate it on the work your team needs to finish, the corrections counsel must make, and the total cost.
Astra for Law's legal search and specialist configuration also warrant evaluation when your team can access them.
Start with one recurring matter: a vendor agreement, its amendments, and your approved positions. Run the research, review, and redline your team needs to finish. Compare the findings, the corrections counsel must make, and the total cost.
**Start your 14-day free trial or install the GC AI ChatGPT plugin.**
Frequently Asked Questions
What Is Astra for Law?
Astra for Law is OpenAI's AI foundation for law firms and legal technology companies. It combines GPT-6 Astra with settings, tools, and context tailored for professional legal work, including a legal search index and instructions for legal analysis and writing. The legal search index covers U.S. case law, statutes, regulations, court rules, and administrative decisions.
How Accurate Is Astra for Law?
OpenAI's evaluation found 54.0% overall correctness for Astra for Law versus 38.7% for base GPT-6 Astra with web search on 200 U.S. legal research questions from the private validation set of Vals AI's Legal Research Bench. The result measures the complete configured system at the highest reasoning effort, not the model alone. It is a benchmark result, not a production accuracy guarantee or a GC AI comparison. Test representative work and review the sources and corrections before relying on an output.
Is Astra for Law Available to In-House Counsel?
Astra for Law itself is initially available to selected law firms through Trusted Access in ChatGPT and Codex, and API access is coming soon. The GC AI ChatGPT Plugin is a separate partner integration: if it is available to your workspace, install it and connect GC AI to use GC AI's legal intelligence in ChatGPT. Plugin access does not establish direct access to Astra for Law, so confirm which model and tools the plugin uses before planning a deployment.
How Much Does Astra for Law Cost?
The Astra for Law launch page and law offering page do not list a price. Base-model API rates are not a quote for the legal offering. Compare usage, connected services, lawyer correction time, and implementation alongside any subscription. GC AI's pricing page provides a separate reference point; confirm scope, limits, overages, and Contract Intelligence in any proposal.
