This dataset contains 1,760,000+ Exhibit 10 contract documents extracted from SEC EDGAR filings, covering all material agreements publicly filed by U.S. and foreign registrants from January 1994 through the present. Each record is the full unmodified text of one Exhibit 10 attachment. The dataset is updated daily as new Exhibit 10 documents are accepted by EDGAR. Total compressed size is 33.5 GB, delivered as ZIP archives.
Programmatically retrieve the full list of dataset archive files, download URLs and dataset metadata.
Dataset Index JSON API
Download the entire dataset as a single archive file.
Download Entire Dataset:
Download a single container file (e.g. monthly archive) from the dataset.
Download Single Container:
The dataset includes every document filed under an Exhibit 10.x designation in EDGAR filings: employment agreements, credit facilities, licensing agreements, merger and acquisition contracts, supply agreements, real estate leases, equity compensation plans, stockholder rights agreements, government contracts, and settlement agreements in material litigation, among others. These are the verbatim contract documents that public companies are legally required to disclose under SEC regulations — not management summaries or descriptions of them.
For the form types specified in the exhibit table of Regulation S-K, Item 601, Exhibit 10 attachments are material contracts as defined and required by that rule. These form types are:
10-12B, 10-12G, 10-D, 10-K, 10-Q, 8-K, S-1, S-3, S-4, S-8, S-11, SF-1, SF-3, F-1, F-3, F-4
For all other form types not enumerated in the Item 601 exhibit table — including 20-F, 40-F, proxy statements, tender offer schedules, and investment company forms — registrants also routinely attach material contracts under the Exhibit 10 designation in analogy to the forms explicitly covered by Regulation S-K. A single parent filing may generate multiple records — one per Exhibit 10.x attachment (10.1, 10.2, 10.3, etc.).
Each Exhibit 10 file contains the original content of the document exactly as published by the filer, without any modifications, normalization, or reformatting. The text reflects what EDGAR received and accepted from the registrant. For form types covered by Regulation S-K Item 601, the documents are material contracts. For form types outside the Item 601 exhibit table, material contracts are still routinely included, though the designation may also cover other document types — for example, letters of transmittal in Schedule TO tender offer filings, custodian and investment management agreements in Form N-2 investment company registrations, and government concession agreements or regulatory authorizations in Form 20-F filings by foreign private issuers.
Executive compensation: Employment agreements with base salary, target bonus, equity grant schedule, severance, and post-employment restrictions; change-in-control agreements (single- and double-trigger); RSU, stock option, and PSU award agreements; executive severance plans; deferred compensation plans (409A-compliant); director compensation policies; officer and director indemnification agreements.
Debt and credit arrangements: Revolving credit agreements, term loan A and B facilities, bridge loans, note purchase agreements, guarantee and security agreements, pledge agreements, intercreditor agreements, subordination agreements.
Commercial and operational contracts: Exclusive and non-exclusive licensing agreements, patent licenses, technology transfer agreements, drug development collaboration and co-promotion agreements, supply and master purchase agreements, distribution and reseller agreements, joint venture agreements, franchise agreements.
Transactions and corporate restructurings: Merger agreements, asset and stock purchase agreements, stockholder rights agreements (poison pills), registration rights agreements, voting and support agreements, lock-up agreements, subscription agreements, convertible note agreements.
Real estate: Material facility and headquarters leases, sale-leaseback agreements, ground leases.
Government and regulatory: Material government contracts, settlement agreements in material litigation.
Exhibit 10 is the SEC's designated exhibit category for material contracts, governed by Regulation S-K, Item 601(b)(10). The rule creates three disclosure obligations:
The third category is why virtually every public company's executive employment agreements, equity incentive plans, annual bonus plans, deferred compensation arrangements, and director compensation policies are present in the dataset. Form 8-K, Item 1.01 (entry into or material amendment of a material definitive agreement) and Form 8-K, Item 5.02 (executive appointments and compensation changes) are the primary event-driven triggers generating new Exhibit 10 filings between periodic reports. Beyond the forms explicitly enumerated in Regulation S-K, registrants filing many other form types also routinely attach material contracts as Exhibit 10 — making it the universal standard designation for material contracts across the full EDGAR filing universe.
The Exhibit 10 obligation attaches to any registrant subject to Exchange Act reporting or filing a Securities Act registration statement:
Form 8-K, Item 1.01: The largest real-time source of new Exhibit 10 filings. Four-business-day deadline from contract execution. Covers new credit facilities, M&A signings, licensing deals, partnership agreements, and material contract amendments. A credit agreement signed Monday typically appears on EDGAR by Friday.
Form 8-K, Item 5.02: Executive appointment and compensation filings. Employment agreements for newly appointed CEOs, CFOs, and other named executive officers are filed within four business days.
Form 10-K: Annual consolidation. The exhibit index must list all material contracts in effect — newly entered, materially amended, or carried by reference from prior filings. Large accelerated filers file within 60 days of fiscal year end; non-accelerated filers within 90 days.
Form 10-Q: Mid-year contracts. Material agreements entered during the quarter that were not previously filed appear as Exhibit 10 attachments. Large accelerated filers file within 40 days of quarter end.
Form S-1: Timeline determined by IPO preparation and SEC review cycles (typically 30–90 days per review round). S-1 Exhibit 10 filings introduce the company's full contract inventory to EDGAR.
Healthcare and life sciences companies are the heaviest filers of licensing and collaboration agreements. Leveraged finance and private equity activity drives the volume of credit agreement filings. Technology companies contribute software licensing, cloud services master agreements, and IP cross-licensing. Energy companies file pipeline capacity, power purchase, and offtake contracts. REITs file ground leases, master leases, and property management agreements.
Exhibit 2 is designated for primary acquisition plan documents: the merger agreement or purchase agreement in certain M&A transactions. There is deliberate overlap — the same merger agreement may appear as Exhibit 2 in one filing and as Exhibit 10.1 in another depending on counsel's practice. What is consistent: ancillary transaction documents (voting agreements, support agreements, rollover commitments) are nearly always Exhibit 10 regardless of where the merger agreement appears. Full M&A document analysis requires both exhibit datasets.
Exhibit 4 covers instruments defining the rights of security holders: indentures, warrant agreements, deposit agreements for ADRs. An indenture for senior notes appears as Exhibit 4; the related intercreditor agreement, security agreement, and subsidiary guarantees typically appear as Exhibit 10 in the same filing. Stockholder rights agreements historically migrated between Exhibit 4 (older filings) and Exhibit 10 (newer filings); comprehensive poison pill research requires checking both categories for pre-2005 filings.
Exhibit 21 is a list of subsidiary names and jurisdictions — no contract content, no overlap in analytical purpose. It is useful contextually for identifying entities that appear as guarantors or counterparties in Exhibit 10 agreements.
Exhibit 99 is a catch-all for press releases, presentations, and miscellaneous documents. When a material agreement is announced in a press release (Exhibit 99.1), the operative contract appears as Exhibit 10.1 in the same 8-K. Exhibit 99 describes the event; Exhibit 10 contains the agreement.
The 10-K filing dataset contains management's descriptions of material contracts in Item 1 (Business) and Item 7 (MD&A). These descriptions are useful for context and summaries of economic terms, but they are management-authored characterizations, not the contract itself. The Exhibit 10 dataset provides the underlying document. The 10-K body tells you why a contract matters; the Exhibit 10 document tells you what it says. The two datasets are complementary.
Proxy statements contain structured compensation tables and narrative summaries of executive pay arrangements under Item 402 of Regulation S-K. These are required summaries of the employment agreements and compensation plans that appear as Exhibit 10. Proxy data provides organized, field-level compensation figures; Exhibit 10 provides the full legal text with operative provisions, vesting mechanics, restrictive covenant language, and termination triggers. Both are needed for complete compensation analysis.
Products such as Kira, Luminance, and Evisort extract and parse SEC-filed contracts, offering clause-level search and obligation tagging as a service. These are analytics layers built on top of EDGAR's raw document corpus — the same corpus that constitutes this dataset. Users who need clause extraction as a managed service use commercial platforms; users building their own NLP pipelines, training legal language models, conducting large-scale empirical research, or requiring the full document corpus work directly with the raw dataset.
A dataset built from 8-K filings captures event-driven contract execution but misses contracts first disclosed in S-1 registration statements. A 10-K dataset captures annual summaries but misses same-day contract filings. The Exhibit 10 dataset covers all parent form types — making it the only single source that captures a contract's full EDGAR presence from its S-1 introduction through its 8-K execution filing through its 10-K annual inclusion.
Corporate attorneys and in-house counsel benchmark contract terms against comparable filed agreements when drafting or negotiating — severance multiples in executive employment contracts, covenant levels in credit agreements, royalty rate ranges in licensing deals, and non-compete durations in supply and acquisition agreements. Litigators and regulatory counsel retrieve the EDGAR-filed version of a disputed contract, using the accession number and filing date as a timestamp. Attorneys advising on confidential treatment requests review peer company redaction patterns in comparable filed agreements to calibrate their own omissions under the post-2019 standard.
Investment bankers and M&A advisors analyze filed merger agreements to benchmark MAC definitions, termination fee structures (market range: 2–4% of deal value), reverse termination fees, and closing conditions. Leveraged finance teams review filed credit agreements to identify prevailing covenant structures, EBITDA add-back practices, and basket mechanics across the credit cycle.
Private equity and hedge fund professionals pull Exhibit 10 filings for acquisition targets to map contractual obligations — credit covenant headroom, change-of-control assignment restrictions, customer contract renewal windows. Portfolio monitoring teams track credit agreement amendments to detect covenant stress. Activist funds review employment agreements, severance plans, and stockholder rights agreements to quantify golden parachute exposure and map anti-takeover mechanisms.
Buy-side and sell-side analysts read filed licensing and customer contracts to construct financial models for revenue-concentrated companies. Credit analysts at rating agencies review credit agreement covenant packages — maintenance tests, restricted payment baskets, incremental facility mechanics — beyond what management summarizes in earnings commentary. Governance researchers extract executive compensation contracts to analyze clawback provisions, pay-for-performance structures, and non-compete terms at scale.
Legal AI and NLP researchers use Exhibit 10 filings as training and evaluation data for contract NLP tasks. The CUAD dataset (Contract Understanding Atticus Dataset) — a widely used legal NLP benchmark covering 41 clause types across ~510 annotated contracts — was sourced from EDGAR Exhibit 10 filings. Companies building contract review, clause extraction, obligation detection, and due diligence automation tools train and validate models on this corpus.
Empirical finance and law researchers use the corpus for studies on covenant tightness and financial distress prediction, executive severance and managerial risk-taking, MAC clause drafting and litigation outcomes, pharmaceutical licensing structure and development outcomes, non-compete prevalence and labor market effects, and SPAC governance design.
Compliance officers and regulatory staff use Exhibit 10 filings of counterparties and acquisition targets to identify change-of-control clauses and assignment restrictions before transactions close. SEC Division of Corporation Finance staff reference Exhibit 10 filings in comment letter reviews; DOJ and FTC antitrust reviewers examine competitor commercial agreements — exclusive dealing provisions, MFN clauses, geographic restrictions — during merger reviews.
Executive Compensation Benchmarking: Compensation committees and HR executives query Exhibit 10 employment agreements filed by peer companies — same SIC code, comparable revenue and market cap — to extract base salary, target bonus percentages, equity vesting schedules, severance multiples, and CIC trigger structures. Published compensation surveys lag 12–18 months; filed agreements reflect actual negotiated terms at a known date. A technology company hiring a new CFO, for example, pulls 8-K/Item 5.02 filings from 25 comparable software companies to establish a market median before structuring the offer.
Credit Agreement Covenant Benchmarking for Leveraged Finance: Leveraged finance bankers and direct lenders extract filed credit agreements from comparable LBO transactions to identify prevailing covenant levels — maximum net leverage ratios, EBITDA add-back caps, restricted payment basket sizes, and incremental facility accordion mechanics. A PE fund acquiring a healthcare services business at 6.0x EBITDA, for example, pulls term loan B credit agreement filings from comparable SIC codes over the prior 18 months to inform the covenant grid for the new facility.
MAC Clause Analysis for M&A Transactions: M&A attorneys and deal arbitrageurs extract and compare MAC/MAE definitions from filed merger agreements to track how market practice evolves across economic exclusion carve-outs, pandemic event language, and the disproportionate impact standard. Following a Delaware Chancery Court ruling on carve-out scope, a law firm can survey S-4 and 8-K merger agreement filings over the subsequent 12 months to map how post-decision drafting changed and update its standard-form clause accordingly.
Pharmaceutical Licensing Agreement Research: Life sciences analysts and biotech business development professionals search executed collaboration agreements to extract upfront fees, milestone schedules, royalty rate tiers, sublicensing economics, and stage-gate termination rights. A biotech negotiating a co-development agreement for a Phase 2 oncology asset can identify six comparable Exhibit 10 filings from the prior three years and calibrate deal terms against the observed upfront and total milestone ranges.
NLP Model Training for Contract Clause Extraction: AI/ML engineers download the corpus to build training datasets for legal NLP tasks — clause boundary detection, clause type classification, named entity extraction, obligation identification, and risk clause flagging. A legal AI startup building an M&A due diligence tool can create a stratified training corpus of 50,000 Exhibit 10 documents, label change-of-control and assignment restriction clauses, and fine-tune a legal language model against the CUAD benchmark.
Real-Time Material Contract Monitoring: Investment bank deal teams and hedge fund analysts configure daily pulls of new 8-K/Item 1.01 Exhibit 10 filings across sector watchlists to retrieve the full text of new credit facilities, strategic partnership agreements, and merger signings within hours of EDGAR acceptance. A healthcare sector fund monitoring 200 companies can have the full text of a definitive merger agreement — MAC definition, financing condition, termination fee, outside date — before sell-side research begins circulating.
Stockholder Rights Agreement Analysis for Activist Investing: Activist funds and governance researchers retrieve poison pill documents to extract trigger thresholds (typically 10%–20%), permitted bidder exception requirements, sunset provisions, and board exemption discretion, with daily 8-K monitoring providing immediate notice of new adoptions. An activist fund accumulating shares near the trigger threshold can retrieve a newly filed rights agreement, confirm the beneficial ownership trigger and term structure, and adjust its accumulation schedule accordingly.
The dataset includes the full text of all Exhibit 10 attachments extracted from EDGAR filings from January 1, 1994 through the present date, updated daily. Coverage of large-cap domestic companies is comprehensive from 1994; mid-cap and smaller company coverage densifies after mandatory EDGAR filing rules expanded through 1996–2000. Foreign private issuers joined mandatory EDGAR filing on a phased schedule between 1996 and 2002.
File organization: Files are stored in per-month ZIP containers with per-filing subfolders named by accession number (hyphens removed). Later filings preserve original registrant-assigned filenames. Early-era filings where exhibits were extracted from SGML complete submission text files use incremental filenames: exhibit10-1.txt, exhibit10-2.txt, etc. Example: 0001234567240000001/exhibit10-1.txt.
No embedded metadata: The stored exhibit files do not contain filing metadata (accession number, CIK, form type, filed date, exhibit label). Accession number, CIK, form type, filed date, and exhibit label can be retrieved via the Filing Query API.
Confidential redactions: Commercial terms, pricing, royalty rates, and customer names are frequently redacted. Pre-2019 formal CT-order redactions replaced entire sections with blank space or explicit omission statements. Post-2019 inline redactions use bracketed placeholders such as "[***]". The omitted text is not present in the filed document and cannot be recovered from the dataset.
Incorporation by reference: When a company incorporates a previously filed Exhibit 10 by reference in a subsequent filing without re-filing the document, no new record is created. The dataset contains only the original file instance. This is standard practice for unchanged contracts carried forward across annual filings.
Amendment records: A 10-K/A or 8-K/A that re-files a revised contract creates a new subfolder with its own accession number. Reconstructing a contract's amendment history requires matching records by company and contract subject matter across accession numbers, not by accession number alone.
Document format evolution: 1994–2004 filings are predominantly ASCII text, many extracted from SGML submission bundles. HTML became dominant by approximately 2005. Post-2009 exhibits may include HTML tables for embedded financial schedules. Text extraction preserves readable prose but collapses table structure.
Does the dataset include the parent filing body (e.g., the 10-K narrative or 8-K body text)? No. The dataset contains only the Exhibit 10 attachment files. The 10-K narrative, 8-K body text, and S-1 prospectus body are not included.
Are exhibits incorporated by reference included? No. When a registrant incorporates a prior Exhibit 10 filing by reference without re-filing the document, no new file is present in EDGAR and no new record exists in the dataset. Only filings where EDGAR received the actual document are represented.
Does the dataset include Exhibit 10 documents from all SEC form types? Yes. The dataset spans all parent form types that generate Exhibit 10 attachments — 10-K, 10-Q, 8-K, S-1, S-4, 20-F, and all others — rather than being limited to one form type.
Does the dataset include contracts with redacted provisions? Yes. The filed version of the document is included, with redacted sections appearing as blank space, explicit omission statements, or bracketed placeholders depending on the redaction method. The omitted text is not recoverable.
What is the difference between the early-era and later-era file formats?
For filings from approximately 2000 onward, where exhibits were submitted as standalone files, the original EDGAR filename is preserved in the dataset (e.g., ex101-creditagreement.htm). For earlier filings where EDGAR received a single SGML complete submission text file, each Exhibit 10 section is extracted and saved as a sequentially-named plain text file: exhibit10-1.txt, exhibit10-2.txt, etc. Both types appear as standalone files inside per-filing subfolders within the ZIP container.
How does this dataset relate to Exhibit 2 and Exhibit 4 datasets? Exhibit 2 covers primary acquisition plan documents (merger agreements, purchase agreements in specific contexts). Exhibit 4 covers instruments defining security holder rights (indentures, warrants). Exhibit 10 captures the broader category of material commercial contracts not assigned to those designations — including ancillary M&A documents, credit agreement components, licensing arrangements, and all compensatory plans. For M&A analysis, both Exhibit 2 and Exhibit 10 datasets are needed. For credit structure analysis, both Exhibit 4 and Exhibit 10 datasets are relevant.
What is the CUAD dataset and how does it relate? The Contract Understanding Atticus Dataset (CUAD) is the most widely cited legal NLP benchmark, containing approximately 510 commercial contracts annotated with 41 clause types. All CUAD documents were sourced from EDGAR Exhibit 10 filings. This dataset is the larger source corpus from which CUAD drew its sample — providing the full 1.76M+ document corpus for model training beyond the CUAD benchmark sample.