By Patrick Turcotte, COO, Docutrax. Last reviewed: 11 August 2026.

A hold harmless agreement is a contract provision in which one party agrees to absorb another party’s liability for injury, damage, or loss arising out of the work or relationship between them. In construction and vendor contracts, the liability being absorbed is almost always liability owed to someone outside the contract. A subcontractor’s employee is injured on site, sues the property owner, and the clause obligates the subcontractor to take on the owner’s exposure to that claim.

That third-party mechanic is the point of the instrument, and it is worth stating plainly, because the clause is often described as a promise not to sue. A promise not to sue is a different instrument doing a different job. The injured worker never signed the contract.

The clause shifts risk on paper. Whether the shift holds depends on the state law governing the contract, how the clause is worded, and whether there is insurance behind it that will actually respond. That last part is the one most often assumed rather than checked.

What the clause does, and what it gets confused with

Two instruments end up filed under the same heading.

An exculpatory clause, or a covenant not to sue, is a first-party waiver. One party gives up its own claims against the other. It runs between the two signatures on the page and reaches nothing beyond them.

Indemnification is a third-party instrument. The indemnitor agrees to answer for the indemnitee’s liability to someone else: an injured worker, an adjoining owner, a member of the public. In New York, where Labor Law section 240 imposes absolute liability on owners and general contractors for elevation-related injuries and section 241(6) imposes a non-delegable duty tied to specific Industrial Code violations, this is the mechanism carrying most of the exposure on a construction project, because under section 240 the owner’s liability to the injured worker attaches regardless of whether the owner did anything wrong.

Construction and vendor contracts need the second instrument. A clause drafted as the first will not do the work anyone expects of it.

Hold harmless clause vs indemnification agreement

In most jurisdictions, courts and the standard legal dictionaries treat the two terms as functionally synonymous. Some drafters argue that “hold harmless” reaches liability itself while “indemnify” reaches only losses actually paid out. The argument is defensible in the abstract and rarely decides anything.

The distinction that does decide things is indemnify versus defend. The duty to defend is the broader promise and it arrives earlier. In most states it is triggered by what a complaint alleges, before anyone has established fault, and it can obligate the indemnitor to fund the defense of a claim that turns out to be meritless, though some states limit the defense obligation to the indemnitor’s own share of fault. The duty to indemnify attaches to liability once liability is determined. A clause requiring a third party to indemnify but silent on defense can leave the organization funding its own defense from the day the complaint lands, which in a construction claim is frequently the larger number, though some states read a defense obligation into an indemnity promise unless the contract says otherwise.

This is why most construction and vendor contracts use the combined “indemnify, defend, and hold harmless” formulation rather than choosing between the terms.

The three forms of indemnity, and why the difference matters

A fourth failure mode sits outside the table. A clause can be entirely enforceable in its jurisdiction and still be uninsurable, because the coverage that would have funded it was narrowed by endorsement. Enforceability and insurability are separate questions answered by separate documents, and the second one is covered further down.

Why state law can override the clause entirely

Most states have passed anti-indemnity statutes covering construction contracts, and they do not all draw the line in the same place. A September 2021 industry analysis counted roughly 15 states restricting indemnification only where the indemnitee’s own sole negligence caused the loss, and around 28 states going further to bar indemnification for both sole and partial negligence on the indemnitee’s part (CLM Magazine). Different legal summaries categorize the counts differently depending on how they treat states with narrower carve-outs, so treat any specific number, including this one, as a starting point for review by counsel rather than a final answer for a given contract’s jurisdiction.

Texas adds a separate requirement on top of its statute. Under the state’s fair notice doctrine, a clause covering the indemnitee’s own negligence has to satisfy two tests. The express negligence rule requires that the intent to indemnify the indemnitee for its own negligence be stated explicitly within the document rather than implied by general language. The conspicuousness requirement is separate, and asks whether the clause is presented in a way that would attract a reasonable person’s attention, which is why these provisions so often appear in bold or capitals. Both have to be satisfied, and a clause that fails either one will not reach the indemnitee’s own negligence.

New York is worth naming for a different reason. General Obligations Law 5-322.1 voids an agreement to indemnify a party for its own negligence in a construction contract. In Brooks v. Judlau Contracting, Inc., 11 N.Y.3d 204 (2008), the New York Court of Appeals held that a clause qualified by “to the fullest extent permitted by law” limits rather than expands the promisor’s obligation, so a partially negligent general contractor can still recover from its subcontractor for the share of fault attributable to that subcontractor. That savings language is why nearly every New York construction indemnity clause opens with those words, and its absence from a New York contract is worth catching before signature rather than after a claim.

The practical result: a hold harmless clause that looks airtight on the page can be unenforceable the moment a dispute lands in the wrong state, or in the right state with the wrong wording.

Some anti-indemnity statutes reach the insurance too

A common assumption is that where a statute restricts the indemnity clause, additional insured coverage can carry the risk transfer instead. In some states it can. In others it cannot, because the anti-indemnity statute expressly voids requirements to procure additional insured coverage for the indemnitee’s own negligence, on the reasoning that this is the same transfer routed through a different door. Texas took that approach in its anti-indemnity legislation, enacted in 2011 and effective for construction contracts entered into from January 1, 2012, which voids additional insured coverage to the same extent as the indemnity provision it supports.
Other states preserve the insurance route. New York courts have long treated an agreement to procure insurance as a distinct obligation from an agreement to indemnify, so the insurance requirement can survive in circumstances where the indemnity clause would be void.
Which side of that line a state falls on changes the entire risk transfer approach for contracts there. It is a question for counsel in the governing jurisdiction, not something to carry over from a template drafted for a different state.

The coverage that actually stands behind the clause

If the clause is only as good as the insurance behind it, the insurance behind it needs a name. It is not the additional insured endorsement. Additional insured status and contractual indemnity are two separate risk transfer mechanisms that happen to live in the same contract.
What funds an indemnity obligation is the indemnitor’s own commercial general liability policy, through contractual liability coverage. The mechanism runs in two moves. The standard CGL excludes liability the insured assumes under a contract, apart from liability it would have had anyway without the contract. It then gives that exclusion back for an “insured contract,” a defined term that includes the part of a contract in which the insured assumes the tort liability of another party to pay for bodily injury or property damage to a third person. That exception is what makes the third party’s promise collectible from an insurer rather than dependent on the third party’s balance sheet.

It is also exactly where coverage gets narrowed quietly. The ISO CG 24 26 amends the insured contract definition so that assumed tort liability is covered only where the injury or damage was caused, at least in part, by the named insured or those acting on its behalf, which pulls the coverage back from broad form to something closer to intermediate form. The ISO CG 21 39 goes further and strikes that part of the definition altogether, and manuscript endorsements can do the same thing in a carrier’s own wording. The employer’s liability exclusion carries its own insured contract exception, and on New York work in particular, action over and employee injury exclusions are written to strip it out, which closes the route a Labor Law claim takes back to the subcontractor’s policy. A third party can sign a broad indemnity clause, the clause can be fully enforceable in that state, and the coverage standing behind it can still be hollow because a single endorsement changed a definition inside the policy.

None of that appears on a certificate. It appears in the policy’s definitions and its endorsement schedule.

An indemnity clause does not create additional insured status

The indemnity provision and the insurance procurement provision do separate jobs, and a contract needs both.
Blanket additional insured endorsements are triggered by a written contract requiring the named insured to procure additional insured coverage. An indemnity clause is not that requirement. It obligates the third party to answer for liability. It does not obligate the third party to add anyone to its policy.

Gilbane Bldg. Co./TDX Constr. Corp. v. St. Paul Fire & Marine Ins. Co., 31 N.Y.3d 131 (New York Court of Appeals, 2018) shows how narrowly those trigger conditions can be read. The endorsement extended additional insured status to organizations the named insured had agreed in a written contract to add. Gilbane, the construction manager, held a contract with the project owner rather than with the named insured, and the court held that this did not bring it inside the endorsement. Reviewing the case, one law firm’s analysis put the practical lesson plainly: a standard indemnification or hold harmless clause will not suffice (Smith, Gambrell & Russell).

So a contract with an airtight hold harmless clause and no insurance procurement requirement can leave an organization holding a promise and no additional insured status, because the endorsement’s trigger condition was never met.

What to check before relying on one

1. Identify which form of indemnity the clause actually uses (broad, intermediate, or limited). The wording determines that, not the heading above it.

2. Check the governing state’s anti-indemnity statute, or have counsel check it, rather than assuming a clause drafted for one jurisdiction survives in another.

3. Confirm the clause meets any additional state-specific requirements, such as the two prongs of Texas’s fair notice doctrine or New York’s savings language.

4. Confirm the contract separately requires the third party to procure additional insured coverage, and states the scope of it, including completed operations where the nature of the work warrants that. The indemnity clause does not do this on its own.

5. Verify the insurance standing behind the promise. That means requesting the specific endorsements from the third party’s brokers and agents rather than relying on the certificate, which carries the limitation stated in ACORD’s own disclaimer: it does not amend or alter the coverage the policy provides. Where exposure warrants going further, it also means reading the policy’s insured contract definition and endorsement schedule, since that is where contractual liability coverage is narrowed or removed.

Step five is where certificate plus endorsement review earns its place, and where reading the policy itself becomes the proportionate step for higher-exposure relationships. A hold harmless clause records what a third party has agreed to do. The endorsements and the policy language are what show whether an insurer stands behind it.