Harvey emerged from stealth on November 23, 2022, announcing a $5 million funding round led by the OpenAI Startup Fund. Jeff Dean, Elad Gil and other angel investors also participated. The funding was for an early legal-AI company founded by Winston Weinberg and Gabriel Pereyra—not a promise that an OpenAI chatbot could replace lawyers, and not necessarily a direct investment from OpenAI’s operating company.
The announcement described Harvey as a natural-language “copilot for lawyers.” Its significance is clearer in hindsight: the narrow beta became a much broader enterprise platform for legal, tax and financial work, while the core risks of hallucinated authority, confidentiality and attorney oversight remained central.
What happened in November 2022?
Harvey announced its emergence from stealth and a $5 million financing round on November 23, 2022. The round was led by the OpenAI Startup Fund, with Google AI leader Jeff Dean, Elad Gil and other angels joining.
TechCrunch described Weinberg as a former securities and antitrust litigator at O’Melveny & Myers and Pereyra as a researcher associated with DeepMind, Google Brain and Meta AI. That combination—legal practice and machine-learning research—shaped Harvey’s original pitch.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors#1 Best Overall
What Harvey’s first product did
The launch product was a natural-language interface for legal work, not a consumer legal-advice service. Users could describe a task in ordinary language and receive research, analysis or draft text.
- Ask questions about case law and legal rules.
- Compare legal concepts, such as whether a worker is an employee or an independent contractor.
- Review or rewrite contract clauses, including a lease provision under California law.
- Draft legal arguments and other documents.
- Turn plain-English instructions into outputs that fit a legal workflow.
Harvey’s launch disclaimer said the product was not intended to provide legal advice to nonlawyers and should be used under the supervision of a licensed attorney. That distinction matters: drafting assistance and research support are not the same as advising a client, filing a pleading or exercising professional judgment.
What “funded by OpenAI” means here
The precise description is that Harvey raised $5 million in a round led by the OpenAI Startup Fund. The fund was presented at the time as a vehicle for investing in early-stage AI companies, with portfolio companies receiving capital and access to new OpenAI systems and Microsoft Azure resources.
That wording is more accurate than saying OpenAI, Inc. simply wrote a corporate check or acquired Harvey. The cited announcement establishes the fund’s role, but it does not establish every detail of the fund’s legal ownership or economics.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Why OpenAI backed a legal-AI startup
The investment reflected two related ideas. First, Harvey was an early example of applying large language models to a high-value professional domain. Legal work contains repeatable research and drafting tasks, but it also rewards domain context, source control and workflow integration.
Second, OpenAI executive Brad Lightcap said Harvey’s vision could help lawyers deliver better services more efficiently and reach more clients. That is an access-and-productivity thesis, not evidence that generated legal answers are automatically reliable.
Why legal AI is unusually difficult
Hallucinated authority
A model can produce a convincing but nonexistent case, statute, quotation or citation. In legal practice, that can create professional-discipline, financial and reputational exposure. Every authority must be checked against the underlying source and the governing jurisdiction.
Confidentiality and privilege
Lawyers may upload privileged, confidential or commercially sensitive material. Buyers should ask whether customer data trains general models, how long prompts and files are retained, whether deletion is available, how tenants are isolated, where data is stored, and what contractual and access-control protections apply.
Recommended Free Tools
At launch, Harvey told TechCrunch that it anonymized data, deleted it after predetermined periods, accepted deletion requests and did not cross-contaminate data between clients. Those were company claims about the early product; they should not be assumed to describe every current configuration without reviewing current contracts and documentation.
Reliance and unauthorized-practice boundaries
Harvey was positioned for attorney-supervised professional work. A consumer should not treat it as a substitute for a lawyer, and a firm should define who approves research, drafts, filings and client-facing advice.
Jurisdiction, currency and missing facts
Even a strong answer can fail when a statute has changed, a local rule controls, the document is poorly scanned, authorities conflict, relevant facts are missing or the question requires procedural judgment. A benchmark on curated prompts cannot establish universal reliability.
What happened after the launch
| Date | Event | How to read it |
|---|---|---|
| 2022 | Founded by Winston Weinberg and Gabriel Pereyra. | Founder backgrounds reported by TechCrunch and OpenAI. |
| November 23, 2022 | $5 million round led by the OpenAI Startup Fund. | The funding announcement described by the original headline. |
| 2023–2024 | $80 million Series B at a $715 million valuation; development of a custom case-law model with OpenAI. | Figures and performance claims come from OpenAI’s company profile. |
| February 12, 2025 | $300 million Series D at a $3 billion valuation. | Sequoia led; the OpenAI Startup Fund participated, according to Harvey. |
| December 2025 | TechCrunch reported an $8 billion valuation after an Andreessen Horowitz-led round. | Secondary reporting, distinct from the company announcements. |
| March 25, 2026 | $200 million financing at an $11 billion valuation, co-led by GIC and Sequoia. | Harvey’s announcement does not list OpenAI as a participant in that specific round. |
| August 2026 | More than 2,400 customers in 70-plus countries and more than 75 AmLaw 100 firms. | Self-reported figures on Harvey’s company page. |
How the product evolved
OpenAI later described Harvey as a secure generative-AI platform for law, tax and finance. The companies worked on a custom-trained case-law model that began with Delaware case law and expanded to U.S. case law, using data equivalent to 10 billion tokens, according to OpenAI.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Rank #4
- Author: Guillebeau, Chris.
- Publisher: Currency
- Pages: 304
- Publication Date: 2012-05-08
- Edition: NO-VALUE
OpenAI said attorneys from 10 large law firms tested the custom model against GPT-4. In OpenAI and Harvey’s reported comparison, the custom model produced an 83% increase in factual responses and attorneys preferred its outputs 97% of the time. These are company-reported results, not an independent or peer-reviewed benchmark, and they do not remove the need for source verification.
Harvey’s current company materials describe a broader enterprise platform spanning contract analysis, due diligence, compliance, litigation, AI agents, document and knowledge workflows and legal operations. Those capabilities should not be projected backward onto the 2022 beta.
How to evaluate Harvey for a legal team
- Match the use case. Identify whether the priority is litigation research, contract review, due diligence, transactional drafting, compliance, internal knowledge retrieval or repetitive operations.
- Inspect source grounding. Require inspectable citations and a clear distinction between retrieved authority and generated analysis.
- Set human approval points. Use attorney review, audit logs, version history and explicit responsibility for final work product.
- Review data governance. Confirm retention, deletion, training use, tenant isolation, identity controls, integrations and contractual protections.
- Measure the workflow. Track time per matter, rework, citation-verification time, adoption and error rates—not just chatbot response speed.
- Price the completed work. Enterprise legal-AI products are generally sold through demonstrations and customized contracts. Compare cost per completed matter or workflow rather than an assumed consumer subscription price.
When Harvey may be a poor fit
- A consumer seeking personal legal advice.
- A solo or small practice that needs transparent, low-cost software.
- A firm without an AI-use policy or attorney-review process.
- Matters involving jurisdictions or source collections the system does not adequately support.
- Highly sensitive work that has not passed privilege and security review.
- Teams expecting autonomous legal work with no professional verification.
- Organizations that mainly need a licensed legal-research database rather than broad workflow automation.
Alternatives worth comparing
Legora
Legora presents a collaborative legal-AI platform with agents, research, document integrations, monitoring and security controls. Its public site promotes booking a demo rather than displaying standard pricing. Compare jurisdictional coverage, integrations, governance and implementation support rather than assuming either platform is superior.
Lexis+ with Protégé
LexisNexis says Lexis+ AI became Lexis+ with Protégé in February 2026. The product combines AI with LexisNexis primary and secondary sources, Practical Guidance, drafting, document analysis, workflow automation and Shepard’s citation validation. Pricing varies by organization, capabilities and content access; LexisNexis advertises customized quotes and a two-day free trial on its product page.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Best Value
Westlaw Precision AI
Firms standardized on Thomson Reuters may also evaluate Westlaw Precision AI. Product names, coverage, deployment terms and pricing change, so verify the current official offering before comparing it with Harvey.
General-purpose enterprise AI
ChatGPT Enterprise, Microsoft 365 Copilot and Claude for enterprise may suit general document analysis or internal knowledge work. They can be more flexible, but may lack licensed legal databases, legal-specific citation systems, matter-centric workflows and law-firm implementation support. Compare the complete workflow, source coverage, data controls and human-review burden—not just model quality or monthly price.
The bottom line on the 2022 announcement
Harvey’s November 23, 2022 announcement marked an early, influential attempt to apply foundation models to legal work. The important fact is specific: the OpenAI Startup Fund led a $5 million round. The later story is not simply that a chatbot answered legal questions; it is that Harvey pursued controlled, domain-specific infrastructure for professional workflows while facing the same unresolved requirements for accuracy, confidentiality, governance and attorney judgment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
