INDEPENDENT VERDICT
Harvey AI Corp's September 11, 2026 Form D confirms one of the largest recent legal-technology financings: $550,299,446 of Equity sold to 28 investors under Rule 506(b), with an indefinite total offering and no sales commissions, finder fees or Item 16 related-person payments. Harvey is a Delaware corporation formed in 2022, formerly named Counsel AI Corp, and the SEC filing lists founders Winston Weinberg and Gabriel Pereyra alongside directors Pat Grady and Ilya Fushman and senior executives John LaBarre, Alan Ghelberg and Katie Burke. The SEC number aligns closely with Harvey's September 9 public announcement of a $550 million financing at a $15.5 billion valuation, co-led by Diffusion and Lightspeed Venture Partners; the roughly $299,446 difference is best understood as the distinction between an exact Form D securities-sales figure and a rounded public financing headline, not evidence of a separate hidden round.
The financing matters because Harvey is no longer presenting itself merely as an application layer sitting on top of OpenAI, Anthropic or other frontier models. In August 2026 it introduced Harvey Tenet, its first post-trained open-weight legal model, and its September financing announcement says the new capital will help scale model training to thousands of GPUs, deploy legal foundation models throughout the Harvey platform and allow customers to train organization-specific models that they own. That changes the investment thesis substantially: Harvey is attempting to move vertically down the AI stack from legal workflow software into model development, post-training infrastructure and proprietary institutional intelligence. The $15.5 billion valuation is therefore being placed not only on a fast-growing legal SaaS product but on the premise that law firms and corporate legal departments will increasingly want specialized AI models trained around their own precedent, drafting standards, knowledge and judgment rather than relying entirely on generic frontier-model APIs.
THE SEC NUMBER AND THE $550M HEADLINE ARE ALMOST — BUT NOT EXACTLY — THE SAME
Harvey's public announcement says the company raised `$550M`, while its Form D reports exactly `$550,299,446` sold. This is a useful example of why FilingDossier should preserve primary filing numbers even when a company has already issued a financing press release. The Form D also leaves the total offering indefinite rather than setting it at $550 million, meaning Harvey could legally report additional sales under the same offering framework if more investors enter or related securities are issued later. The current filing shows 28 investors, first sale on August 28 and $0 minimum investment; it does not disclose individual allocations, share price, preferred-stock designation, liquidation preference or percentage ownership sold.
Harvey's public financing disclosure adds information that the SEC notice does not. Diffusion and Lightspeed Venture Partners co-led the financing, while existing investors Sequoia, Kleiner Perkins, Andreessen Horowitz, Coatue, Conviction, Elad Gil, Evantic, GIC, Goldman Sachs Alternatives, Verified Capital and WNDR participated; Sapphire Ventures and Whale Rock were also named among the new investors. Harvey itself states the round values the business at $15.5 billion. Those terms should remain distinct from the SEC facts: the SEC verifies the issuer, security type, amount sold and investor count, while Harvey supplies the valuation and investor identities. The filing itself does not label the financing Series H, and although Lightspeed describes its investment as Series H, the safest SEC-facing title is therefore "SEC Form D Review" rather than relying on a round label that does not appear in the filing.
HARVEY'S STRATEGY HAS SHIFTED FROM LEGAL COPILOT TO OWNED INTELLIGENCE
The most important operating change behind the financing is Harvey's move toward proprietary legal intelligence. The company says 80% of Am Law 100 firms now use Harvey and that its customer base also includes five of the Fortune 10, giving it access to a large volume of sophisticated legal workflows and institutional knowledge. Harvey's platform has expanded beyond basic chat and drafting into Vault document repositories, contract and due-diligence workflows, multi-agent review systems, legal research, organizational knowledge and enterprise integrations. Its September funding announcement frames the next phase around helping legal organizations "build and own their intelligence," which is materially different from simply reselling access to a general-purpose language model.
That strategic shift is technically visible in Harvey's recent product history. The company has published engineering work on rebuilding Playbook Review as a multi-agent system, scaling document processing, operating its own cloud-agent infrastructure and developing Harvey Tenet as a post-trained open-weight model specialized for legal tasks. Harvey is therefore trying to own more of the layers that determine model behavior: legal data preparation, post-training, evaluation, retrieval, agent orchestration and eventually customer-specific model adaptation. This gives Harvey more differentiation and potentially more control over privacy and performance, but it also raises infrastructure costs dramatically. Scaling model training to thousands of GPUs turns Harvey from a relatively asset-light SaaS company into a business that must manage substantial compute expenditure, model-development talent, inference economics and continual benchmarking against frontier providers whose base models continue improving rapidly.
THE PREVIOUS NAME "COUNSEL AI" SHOWS HOW FAR THE COMPANY HAS MOVED
The SEC record preserves `Counsel AI Corp` as Harvey's previous legal name, providing a clean identity chain back to its earliest phase. Harvey was founded in 2022 by Winston Weinberg, a former securities and antitrust lawyer at O'Melveny, and Gabriel Pereyra, whose background included AI research. The original insight was unusually domain-specific: legal work combines large amounts of confidential text, repetitive drafting and research with high requirements for accuracy, professional judgment and data security. Harvey built around those constraints rather than launching as a broad enterprise assistant and then searching for a legal use case.
That domain focus appears to have driven unusually rapid capital formation. Harvey raised progressively larger rounds as law-firm adoption expanded, including a $300 million financing at a $3 billion valuation in early 2025, another $300 million round at approximately $5 billion later that year, subsequent financings that pushed the company toward an $11 billion valuation by early 2026, and now the $550 million round at $15.5 billion. Historical financing figures should not simply be added to the latest Form D as if they were current cash, because earlier rounds financed operations and may include different securities and secondary components. The relevant point is the pace: Harvey has repeatedly repriced its equity upward while continuing to raise nine-figure amounts, creating very high expectations for future revenue, margins and defensibility.
THE BIGGEST MOAT MAY BE CUSTOMER DATA — AND THAT IS ALSO THE BIGGEST GOVERNANCE PROBLEM
Harvey's push toward organization-specific models creates a potentially powerful competitive advantage. Large law firms possess decades of precedent language, negotiation positions, templates, internal memoranda, deal structures and litigation strategy that generic public training data cannot fully replicate. If Harvey can convert those private work products into models or retrieval systems that remain securely controlled by each organization, customer-specific performance could improve over time and switching costs could increase. Lightspeed's investment thesis explicitly emphasizes institutional knowledge, standards and judgment as the next source of AI differentiation rather than simple access to the same frontier models.
The same strategy creates difficult legal and technical issues. Law-firm data can contain attorney-client privileged material, trade secrets, personally identifiable information, confidential transaction documents and information subject to professional-responsibility rules or client outside-counsel guidelines. Customers therefore need precise answers about whether their data trains shared models, how organization-specific model weights are segregated, where inference occurs, who owns derivative model artifacts, how data is deleted and what happens when a law firm ends its contract. Harvey's strategy of letting organizations "own their intelligence" will only become a durable moat if the contractual and technical architecture makes ownership genuinely meaningful rather than a marketing phrase. A legal-AI vendor can achieve strong benchmark performance and still fail procurement if clients cannot verify privacy boundaries, audit trails and model governance.
$15.5B MEANS HARVEY IS NOW BEING VALUED AS INFRASTRUCTURE, NOT A NICHE LEGAL TOOL
The valuation changes the diligence standard. At $15.5 billion, Harvey is no longer priced like an emerging legal software vendor. Investors are implicitly underwriting a much larger outcome: that specialized AI becomes a core operating layer across global professional services and that Harvey captures a meaningful share of that infrastructure. The company's claim that 80% of the Am Law 100 already use its platform provides powerful distribution evidence, but penetration does not reveal contract size, utilization, renewal rates or profitability. One global law firm can represent anything from a limited deployment to thousands of users, and a Fortune 10 customer logo does not disclose annual contract value.
The Form D declines to disclose revenue range, so current revenue cannot be verified through SEC data. Media reports have circulated annualized revenue figures ranging into the hundreds of millions during 2026, but those are not audited SEC disclosures and should not be used as though Form D verified them. What investors need is current ARR, net revenue retention, gross margin after model and inference costs, average contract value, customer concentration, usage growth, compute cost per legal task and the percentage of revenue attributable to law firms versus in-house legal and professional-services organizations. Harvey's business can grow very quickly while model-training and inference costs grow quickly as well, making gross-margin evolution one of the most important metrics behind the valuation.
FINAL ASSESSMENT
Harvey AI's September 2026 Form D is notable because the SEC record and the company's financing announcement line up unusually well. SEC EDGAR reports $550,299,446 of Equity sold to 28 investors, while Harvey publicly announced a rounded $550 million financing at a $15.5 billion valuation. The filing confirms Harvey AI Corp is the operating company, preserves its previous Counsel AI identity and identifies a board and executive group that includes founders Winston Weinberg and Gabriel Pereyra along with major venture-investor representatives.
The more important independent finding is what the capital is financing. Harvey is using the new round to move from being primarily a legal application built around third-party foundation models toward owning more of the specialized model layer itself. Tenet, open-weight post-training, thousands of GPUs, multi-agent legal workflows and customer-specific model training all point toward a strategy in which proprietary legal intelligence becomes the product rather than a thin interface around external APIs. That could increase differentiation and customer lock-in, but it also raises capital intensity and exposes Harvey to direct competition with increasingly capable frontier-model providers.
The core diligence question is therefore no longer simply whether lawyers adopt AI. Harvey has already demonstrated significant adoption. The harder question is whether a $15.5 billion company can convert that distribution into durable model-level economics while preserving privileged data, controlling inference costs and maintaining enough technical differentiation as base models improve. Investors should obtain the preferred-stock terms, fully diluted capitalization, audited revenue and gross-margin history, customer-retention data, GPU commitments, model-training expenses and contractual rules governing ownership of customer-trained intelligence before using the headline valuation as evidence of fundamental value.
Form D is an exempt-offering notice. It is not SEC approval of Harvey AI, its $15.5 billion valuation, its legal models, its customers or any projected investment return.
SEC SNAPSHOT
ISSUER: Harvey AI Corp | PREVIOUS NAME: Counsel AI Corp | CIK: 0001974654 | SEC FILE NO.: 021-597170 | ACCESSION NO.: 0001974654-26-000003 | FILED / EFFECTIVE: September 11, 2026
ENTITY: Delaware Corporation | INCORPORATED: 2022 | PRINCIPAL ADDRESS: 201 Third Street, Suite 500, San Francisco, CA 94103 | PHONE: 415-579-0539
INDUSTRY: Other Technology | EXEMPTION: Regulation D Rule 506(b) | POOLED INVESTMENT FUND: No | INVESTMENT COMPANY ACT EXCLUSION: None claimed
SECURITY: Equity | BUSINESS COMBINATION: No | FIRST SALE: August 28, 2026 | OFFERING DURATION: One year or less
TOTAL OFFERING: Indefinite | AMOUNT SOLD: $550,299,446 | REMAINING: Indefinite | INVESTORS: 28 | MINIMUM INVESTMENT FIELD: $0
SALES COMMISSIONS: $0 | FINDER FEES: $0 | ITEM 16 RELATED-PERSON PAYMENTS: $0 | REVENUE RANGE: Declined to disclose
RELATED PERSONS: Winston Weinberg — Executive Officer / Director | Gabriel Pereyra — Executive Officer / Director | Pat Grady — Director | Ilya Fushman — Director | John LaBarre — Executive Officer / Director | Alan Ghelberg — Executive Officer | Katie Burke — Executive Officer
FORM D SIGNATORY: John LaBarre | TITLE: Chief Legal Officer
PUBLIC FINANCING ANNOUNCEMENT: $550M at a $15.5B valuation | ANNOUNCED: September 9, 2026 | CO-LEADS: Diffusion and Lightspeed Venture Partners.
PUBLICLY IDENTIFIED PARTICIPANTS: Sequoia | Kleiner Perkins | Andreessen Horowitz | Coatue | Conviction | Elad Gil | Evantic | GIC | Goldman Sachs Alternatives | Verified Capital | WNDR | Sapphire Ventures | Whale Rock.
SEC VS. PUBLIC ROUND AMOUNT: SEC amount sold is $550,299,446, approximately $299,446 above the rounded $550M public announcement. The Form D keeps the offering amount indefinite.
CURRENT PUBLIC ADOPTION CLAIM: Harvey states that 80% of Am Law 100 firms use the platform and that customers include five of the Fortune 10. THESE ARE COMPANY-DISCLOSED ADOPTION METRICS, NOT SEC-VERIFIED REVENUE.
MODEL STRATEGY: Harvey Tenet — first post-trained open-weight legal model | organization-specific model training | multi-agent legal workflows | proprietary cloud-agent infrastructure | planned training across thousands of GPUs.
IMPORTANT MODEL-STACK SHIFT: Harvey historically built on frontier AI providers, but the 2026 strategy increasingly includes its own post-trained models and client-owned organization-specific intelligence. This increases potential differentiation while also increasing compute, engineering and infrastructure requirements.
VALUATION: $15.5B according to Harvey's September 2026 financing announcement. The SEC Form D does not determine or verify company valuation.
CURRENT REVENUE: Not disclosed in the Form D. Media-reported ARR figures should remain separate from SEC-verified financial data.
CORE INDEPENDENT FINDING: Harvey's latest SEC filing is not simply evidence of another large legal-AI financing. It marks the point where Harvey is attempting to move deeper into the AI stack — from legal workflow software using external models toward a platform that post-trains its own legal models and lets customers create proprietary institutional intelligence. The $550.299M SEC sale and $15.5B public valuation therefore depend increasingly on Harvey proving that legal-specific models, customer knowledge and agent infrastructure can create durable economic differentiation rather than becoming features of general-purpose frontier AI systems.
Form D is an exempt-offering notice and is not an SEC-issued certificate, approval or endorsement.