NDA Contract Benchmarks 2026: Data on Standard NDA Terms

The clearest public 2026 NDA contract benchmarks show that fixed terms dominate: 78% of the NDAs in Common Paper's dataset had a fixed agreement term, 79% had a fixed confidentiality period, and 75% of fixed confidentiality periods were two years. Separately, TermScout's 2026 Contract Signals Report found that 38% of enterprise NDAs reviewed in Q1 2026 contained at least one AI-related provision. These figures provide useful comparison points, but they do not make any term appropriate for every transaction.

Key Takeaways

• Public Common Paper data shows that 78% of NDAs had fixed agreement terms and 22% had no expiration.
• Seventy-nine percent used a fixed confidentiality period; 21% used an unlimited period.
• Among fixed confidentiality periods, two years represented 75% of the observed positions.
• A one-year agreement term was the single most common agreement-term structure, although Common Paper does not publish its exact percentage.
• Ninety-six percent selected United States governing law, with Delaware appearing in 84% of the NDA population.
• Separate TermScout data shows that 38% of enterprise NDAs reviewed in Q1 2026 contained at least one AI-related provision.
• Prevalence is context, not a legal conclusion or automatic approval rule.

What NDA Contract Benchmarks Tell You

NDA contract benchmarks compare a proposed confidentiality agreement with positions observed in relevant contract populations. They can help Legal and Procurement determine whether a term is common, less common, or simply unsupported by available prevalence data.

The comparison must be precise. The NDA agreement term controls how long the agreement remains active and ordinarily defines the period during which protected disclosures may be made. The confidentiality period controls how long protection obligations survive. Treating those periods as interchangeable can produce an inaccurate review.

This page is deliberately narrower than TermScout's confidentiality clause benchmark research, which examines confidentiality provisions inside broader SaaS agreements. It is also distinct from TermScout's research on NDA standardization without standard forms. This resource focuses specifically on measurable standalone NDA terms and how teams should use those measurements.

Public NDA Benchmark Data

The primary public prevalence source used here is the Common Paper 2026 SaaS Contract Benchmark Report. Its full dataset includes 16,140 signed Cloud Service Agreements, NDAs, and Design Partner Agreements sent by 2,223 unique companies. The 2026 observation period runs from June 2025 through June 2026.

Common Paper reports that 78% of the NDAs in its population used a fixed agreement term, up from 71% over three years. Twenty-two percent had no expiration. A fixed one-year term was the single most common agreement-term structure, although the report does not provide its exact share.

For surviving confidentiality obligations, 79% used a fixed period and 21% used an unlimited period. Among the NDAs with a fixed confidentiality period, 75% used two years. That 75% figure must not be restated as 75% of all NDAs.

Governing-law choices were concentrated in the United States. Common Paper reports United States law in 96% of the NDA population, including Delaware in 84%, California in 6%, and New York in 4%.

NDA Contract Benchmark Table

TermScout Contract Signals

The public benchmarks above are external Common Paper findings. They are not TermScout proprietary data.

Separately, the TermScout 2026 Contract Signals Report found that 38% of enterprise NDAs reviewed in Q1 2026 contained at least one AI-related provision. TermScout describes the report as using aggregated and anonymized data from agreements reviewed through Certify.

This signal measures the presence of AI-related provisions. It does not establish that a particular AI restriction is balanced, enforceable, necessary, or suitable for a specific transaction. It also should not be blended mathematically with Common Paper's figures because the sources cover different populations and methodologies.

The finding matters operationally because NDAs may now address model training, processing systems, retention, downstream access, or other AI-governance topics before a full procurement, security, or technology agreement is negotiated.

Detailed Analysis of Standard NDA Terms

1. NDA Agreement Term

The agreement term determines how long the NDA is active. In the Common Paper framework, only information shared during that period falls within the agreement's protection. A fixed term appeared in 78% of the 2026 NDA population, and a fixed one-year term was the single most common structure.

Reviewers should confirm the start date, expiration mechanics, termination rights, and treatment of information disclosed shortly before expiration. A common term length is a useful reference point, but the expected duration of diligence, partnership discussions, product evaluation, or acquisition activity may justify a different result.

2. Confidentiality Period

The confidentiality period is the time during which the receiving party must continue protecting covered information. Common Paper found fixed periods in 79% of NDAs and unlimited periods in 21%. Among fixed periods, two years accounted for 75%.

Legal and Procurement should verify when the period begins. Measuring from the effective date can produce a different result than measuring from each disclosure or the last disclosure. The team should also determine whether every category of information receives the same period.

3. Trade-Secret Treatment

A fixed period for ordinary confidential information does not necessarily resolve the treatment of trade secrets. Under 18 U.S.C. § 1839, trade-secret status depends in part on reasonable secrecy measures and independent economic value from the information not being generally known or readily ascertainable through proper means.

Teams should determine whether trade secrets receive separate protection tied to their continuing legal status. This is a drafting and risk question, not a conclusion supplied by the two-year prevalence figure.

4. Definition and Exclusions

The definition identifies what information is protected. The exclusions identify information that is public, previously known, independently developed, or properly received from another source. The Common Paper Mutual NDA provides one public reference model for these concepts, but a standard form is not evidence that each position has a particular market-prevalence percentage.

Review whether oral disclosures, unmarked materials, observations, derivatives, notes, and information disclosed before the effective date are covered. Also check whether the receiving party can operationalize any marking or confirmation requirement.

5. Permitted Use and Disclosure

An NDA should connect use of confidential information to a defined purpose and identify the people or entities that may receive it. Review access by employees, professional advisers, contractors, affiliates, financing sources, and potential acquirers. Confirm whether the receiving party remains responsible for recipients and whether disclosure is limited by a need-to-know standard.

6. Return, Destruction, and Retention

Review when information must be returned or destroyed, whether certification is required, and how backups, legal holds, archival systems, and automated retention are handled. A clause that requires complete deletion may conflict with actual system behavior unless it includes carefully defined retention exceptions.

7. AI Governance Language

Because TermScout found AI-related language in 38% of enterprise NDAs reviewed in Q1 2026, teams should check whether the NDA addresses AI use intentionally or through language broad enough to create uncertainty.

Questions may include whether confidential information can be submitted to an AI system, used for model training, retained by a provider, accessed by subprocessors, or used to generate outputs. Teams should also determine whether these subjects belong in the NDA or should be handled in a security addendum, data processing agreement, AI rider, or primary services agreement.

8. Governing Law and Forum

Delaware's 84% share in the Common Paper population is a prevalence finding, not proof that Delaware is the right choice for every NDA. Review the parties' locations, related agreements, dispute strategy, mandatory law, and whether governing law and forum provisions are internally consistent.

What the Benchmarks Mean for Legal and Procurement

• Common does not mean mandatory. A prevalent term can still conflict with internal policy or the information being disclosed.
• Uncommon does not mean unacceptable. A less common position may address a specific operational or regulatory need.
• Reference standards are not prevalence studies. A committee-drafted form can illustrate a balanced structure without proving how often that structure appears in signed agreements.
• Different datasets should remain separate. Common Paper's public findings and TermScout's enterprise AI signal answer different questions.
• The denominator matters. The two-year figure covers fixed confidentiality periods, not every NDA in the public dataset.

What Teams Should Review, Negotiate, Approve, or Escalate

Review

• Agreement term and termination mechanics.
• Confidentiality period and the event that starts it.
• Definition, exclusions, and treatment of oral or unmarked disclosures.
• Permitted use, recipient groups, and downstream responsibility.
• Return, destruction, retention, and backup exceptions.
• AI use, model-training, retention, and provider-access language.
• Governing law, forum, remedies, and alignment with related agreements.

Negotiate

• Language that is materially broader than the actual purpose of the disclosure.
• Time periods that do not match the expected sensitivity or useful life of the information.
• Deletion obligations that cannot be implemented across existing systems.
• AI restrictions or permissions that conflict with approved technology workflows.
• One-sided obligations that are inconsistent with the direction of information flow.

Approve

A team may approve a position when it fits the transaction, internal policy, information sensitivity, operational capabilities, and applicable approval authority. Market alignment can support the decision, but should not replace the organization's playbook or legal judgment.

Escalate

• Trade secrets or highly sensitive technical information without suitable continuing protection.
• Restrictions that interfere with required legal, regulatory, financing, or diligence disclosures.
• Broad AI-training or reuse rights affecting confidential information.
• Residuals, non-solicitation, intellectual-property transfers, audit rights, or commercial restrictions added to a document presented as a routine NDA.
• Terms that conflict with security controls, records-management obligations, or a related commercial agreement.

Why Reviewing One Clause in Isolation Is Not Enough

An NDA's risk allocation emerges from the interaction of its terms. A two-year confidentiality period may operate very differently depending on whether it begins on the effective date or the last disclosure. A broad definition may be moderated by clear exclusions, while a narrower definition may still create substantial operational burdens if use and deletion obligations are unusually strict.

AI language makes these interactions more important. A no-training restriction can depend on definitions of confidential information, permitted systems, subprocessors, retention, and downstream access. Extracting one clause without evaluating those relationships can produce an incomplete benchmark.

A useful review therefore asks three questions: what the complete agreement requires, how its positions compare with relevant market data, and whether the resulting obligations can be approved and performed.

How Certify™ Helps Analyze an NDA

Certify™ is TermScout's contract intelligence platform for analyzing agreements, benchmarking terms against market data, and surfacing contract signals.

For an NDA review, Certify can help Legal and Procurement move from clause identification to a structured decision about which terms may need approval, negotiation, or escalation. Market context can be evaluated alongside the organization's own standards rather than relying only on a counterparty's assertion that its wording is standard.

Validate Market Alignment with TrustMark™ NDA

TrustMark™ NDA provides a separate path for organizations that want to evaluate and demonstrate the market alignment of the NDA they use with counterparties.

TermScout analyzes and benchmarks the submitted NDA to determine whether it qualifies for TrustMark NDA. This path is relevant when the objective is not only to review an agreement internally, but also to provide counterparties with independent evidence that the NDA has been evaluated against real-world market data.

Methodology and Limitations

This benchmark page separates external public data from TermScout proprietary signals.

Public NDA benchmark data: Common Paper's 2026 report covers 16,140 signed agreements sent by 2,223 unique companies across three agreement categories. Its 2026 figures cover June 2025 through June 2026. The public report does not disclose the number of NDAs within the total dataset.

The Common Paper population reflects agreements created through or associated with its standardized contracting ecosystem. It may not represent every industry, geography, company size, bespoke law-firm form, employment NDA, transaction NDA, or heavily negotiated enterprise agreement. Its percentages should be treated as observed prevalence within that population rather than universal market rules.

TermScout Contract Signal: the 38% AI finding applies to enterprise NDAs reviewed in Q1 2026 and is sourced only to the 2026 Contract Signals Report. TermScout describes the underlying information as aggregated and anonymized. The public landing page does not disclose the exact NDA denominator or every coding rule used to classify an AI-related provision.

Reference guidance: the Common Paper Mutual NDA is used to identify concepts teams may review. Its standard terms do not prove the prevalence of a clause unless a separate benchmark supplies that prevalence data.

This article provides market context, not legal advice. Appropriate NDA terms depend on the facts, applicable law, information involved, operational systems, and the parties' risk decisions.

Frequently Asked Questions

What is the standard term for an NDA in 2026?

There is no universal term. In Common Paper's 2026 NDA population, 78% used a fixed agreement term and a fixed one-year term was the single most common structure. The report does not publish the exact percentage represented by one-year terms.

How long do NDA confidentiality obligations usually last?

Common Paper found that 79% of NDAs used a fixed confidentiality period and 21% used an unlimited period. Among the fixed-period NDAs, 75% used two years.

Does the two-year figure cover all NDAs?

No. The 75% figure applies only to NDAs that had a fixed confidentiality period. It should not be described as 75% of every NDA in the dataset.

Are perpetual confidentiality obligations standard?

They were not the majority position in Common Paper's 2026 NDA population. Twenty-one percent used an unlimited confidentiality period. Whether continuing protection is appropriate depends on the information, including whether it remains a legally protected trade secret.

How common is AI governance language in enterprise NDAs?

TermScout's 2026 Contract Signals Report found that 38% of enterprise NDAs reviewed in Q1 2026 contained at least one AI-related provision. This is a separate TermScout signal and is not part of Common Paper's public dataset.

What should Procurement check in an NDA?

Procurement should check the agreement term, confidentiality period, purpose, permitted recipients, return or destruction requirements, operational feasibility, AI and data-use restrictions, governing law, and any non-confidentiality provisions added to the document.

Is a standard NDA template the same as an NDA benchmark?

No. A template is a reference document. A benchmark measures positions observed in a defined contract population. A standard form can inform drafting without proving that each of its terms is prevalent.

How is this page different from confidentiality-clause benchmark research?

This page focuses on standalone NDA agreement terms, surviving confidentiality periods, governing law, and emerging enterprise NDA signals. TermScout's confidentiality benchmark research examines confidentiality provisions in broader SaaS contracts.

Can Certify analyze an NDA?

Yes. Certify analyzes agreements, benchmarks terms against market data, and surfaces contract signals that help Legal and Procurement determine what may need approval, negotiation, or escalation.

What is TrustMark NDA?

TrustMark NDA is TermScout's path for analyzing and benchmarking an organization's NDA to determine whether it qualifies for independent market-alignment validation.

Turn NDA Benchmarks Into an Actionable Review

Benchmarks are most useful when they change a decision. Use public data to challenge unsupported claims about what is standard, then evaluate the complete NDA against the transaction, your playbook, and relevant market agreements.

Analyze Your NDA with Certify™

Validate Your NDA with TrustMark™ NDA