Technology Agreement Checklist: License Scope, Service Levels, Data Rights, Indemnity, and Exit

By Casey Scott McKay · 2026-01-08

Technology agreements fail on eight provisions, and the license grant is not one of them. This checklist runs a subscription or licensing deal in fourteen phases: intake and structure, scope and users, service levels, data rights, indemnity and liability, security and privacy, subprocessors and audit, term and renewal, exit, statements of work and custom development, the AI provisions, adjacent structures, portfolio management, and a twenty-minute triage for agreements that cannot justify full review. Each box gives the reason, the authority where there is one, and the trap. Boxes marked as gates should clear before signature. A worked negotiation and a worked renewal run throughout.

IP and Technology > Information Technology | Checklist | Published 8 January 2026 - Updated 16 May 2026 | Casey Scott McKay - marksy.us

Summary. Technology agreements fail on eight provisions, and the license grant is not one of them. This checklist runs a subscription or licensing deal in fourteen phases: intake and structure, scope and users, service levels, data rights, indemnity and liability, security and privacy, subprocessors and audit, term and renewal, exit, statements of work and custom development, the AI provisions, adjacent structures, portfolio management, and a twenty-minute triage for agreements that cannot justify full review. Each box gives the reason, the authority where there is one, and the trap. Boxes marked as gates should clear before signature. A worked negotiation and a worked renewal run throughout.

Keywords: procurement intake, deal structure, access grant, authorized users, affiliate definition, usage metrics and overage, uptime and exclusions, chronic failure termination, support severity, customer data ownership, usage and aggregated data, model training, IP indemnity carve-out, security super-cap, security addendum, incident notification hours, data processing agreement, subprocessors, renewal cap, exit and export specification

How to use this checklist

| Phase | What it covers | |---|---| | 1 | Intake and structure | | 2 | Scope, users, and metrics | | 3 | Service levels | | 4 | Data rights | | 5 | Indemnity and liability | | 6 | Security and privacy | | 7 | Subprocessors and audit | | 8 | Term and renewal | | 9 | Exit | | 10 | Statements of work | | 11 | The AI provisions | | 12 | Adjacent structures | | 13 | Portfolio management | | 14 | The twenty-minute triage |

Boxes marked [Gate] should clear before signature.

The matters. A manufacturer negotiating a field service platform with six asks and a real alternative; and a company at its fourth renewal of a customer platform it could not leave.

Phase 1. Intake and structure

[ ] Identify the deal structure.
- Why. Subscription, perpetual license, term license self-hosted, managed service, or professional services. The forms differ and so do the risks.
[ ] Identify what data the service will hold.
- Why. Personal data, regulated data, trade secrets, or nothing sensitive. This determines whether the security addendum is the centerpiece or an afterthought.
[ ] Assess the dependency.
- Why. A system of record the business cannot leave requires exit terms; a peripheral tool does not.
[ ] Establish the spend and the term.
[ ] Assess the vendor.
- Why. An established vendor with thousands of customers will not move on architecture. A growth-stage vendor with a large deal in front of it will move on nearly everything.
[ ] [Gate] Establish whether a competitive alternative is real.
- Why. The single largest source of leverage.
[ ] Write a half-page intake sheet and derive the negotiation plan from it.
- Trap. Redlining every clause of a small subscription with an established vendor, which produces a form response and consumes the review budget.

Phase 2. Scope, users, and metrics

[ ] Read the access grant's adjectives.
- Why. Non-exclusive, non-transferable, internal business purposes. Non-transferable defeats assignment in an acquisition without consent; internal business purposes forecloses service-bureau use.
[ ] Understand what a subscription conveys.
- Authority. 17 U.S.C. § 106; 17 U.S.C. § 117; 17 U.S.C. § 109.
- Why. No copy is delivered, so the essential-step and first-sale provisions do not apply. Termination is a loss of access to a running system holding the customer's data.
[ ] [Gate] Test the user definition against how the business works.
- Trap. A named-user agreement for a business whose usage is occasional and broad.
[ ] [Gate] Fix the affiliate definition.
- Trap. The most common source of unexpected true-up invoices, and it is fixed in one sentence at signature.
[ ] Address contractors and outsourced functions.
[ ] Understand the usage metric and who measures it.
[ ] Negotiate the overage mechanic, not the rate.
- Why. Automatic billing at list, throttling, or a good-faith negotiation are very different outcomes.
[ ] Read the restrictions.
- Authority. 17 U.S.C. § 1201; 18 U.S.C. § 1030.
- Why. Benchmarking, competitive use, and reverse engineering, with the interoperability interaction noted where the customer may need it.

Phase 3. Service levels

[ ] Read the uptime number as minutes.
- Why. 99.9 percent is roughly forty-three minutes a month; 99.5 percent is over three and a half hours; 99 percent is more than seven.
[ ] Read the exclusions.
- Trap. Unlimited scheduled maintenance, which defeats the commitment.
[ ] Check the measurement method and whether degradation counts as downtime.
[ ] Do not negotiate larger credits.
- Why. They are the exclusive remedy in nearly every form and are a pricing adjustment rather than compensation.
[ ] [Gate] Negotiate a chronic-failure termination right instead.
- Why. A defined number of breaches in a rolling period, with termination and a pro-rata refund. It converts the service level from a discount into an exit.
[ ] Negotiate support separately.
- Why. Response and resolution targets, severity definitions, and coverage hours. Severity definitions matter most because the vendor usually assigns severity.
[ ] Ask for a deprecation policy where integrations will be built.

Phase 4. Data rights

[ ] [Gate] Confirm the customer owns its data, with the vendor holding a limited license to process it to provide the service.
- Trap. Anything broader transfers an asset for no consideration.
[ ] Scrutinize usage data and telemetry rights.
- Why. Aggregation, de-identification, whether re-association is possible, and whether it may be shared or used to train models.
[ ] Scrutinize the aggregated data clause where the customer's data is competitively sensitive or where aggregation across a small industry permits inference.
[ ] [Gate] Address model training expressly.
- Why. Treat a training permission as a disclosure and evaluate it against the confidentiality classification of the content.
[ ] Limit the feedback clause where the customer's personnel will be deeply engaged in product design.
[ ] Make data location specific rather than aspirational.
[ ] Confirm the confidentiality provisions create a real duty.
- Authority. 18 U.S.C. § 1839; 18 U.S.C. § 1836.

Phase 5. Indemnity and liability

[ ] Confirm the IP indemnity covers patent, copyright, trademark, and trade secret claims.
[ ] Read the exclusions.
- Why. Modifications, combinations, use outside the agreement, and continued use after notice.
[ ] Negotiate the remedy structure.
- Trap. Procure, modify, or terminate and refund. The vendor will choose refund; obtain a transition period before termination becomes the remedy.
[ ] [Gate] Carve the indemnity out of the liability cap.
- Why. An IP indemnity capped at twelve months of fees does not answer a patent claim.
[ ] [Gate] Obtain a security-incident super-cap.
- Why. Breach exposure routinely exceeds twelve months of fees.
[ ] Carve out confidentiality breach and willful misconduct.
[ ] Accept mutual exclusion of consequential damages.
- Why. A vendor's exposure is otherwise unbounded and it will not agree.
[ ] Understand what drives vendor flexibility.
- Why. Deal size and insurability. A cap the vendor's insurance covers is negotiable; unlimited liability is not.

Phase 6. Security and privacy

[ ] Review the security addendum against the customer's own obligations, not as boilerplate.
[ ] [Gate] Set the incident notification deadline in hours.
- Why. The vendor's delay consumes the customer's regulatory clock, and this is a genuine trade because it is operationally meaningful for the vendor.
[ ] Confirm whether the deadline runs from discovery or from confirmation.
[ ] Obtain certification reports annually, with the right to review exceptions.
[ ] Execute a data processing agreement where personal data is involved.
- Authority. Cal. Civ. Code § 1798.140; Cal. Civ. Code § 1798.100.
[ ] Confirm the vendor can delete a named individual's data on request.
- Authority. Cal. Civ. Code § 1798.105.
- Trap. A system that disables an account without removing the record.
[ ] Flow down sector rules.
- Authority. 15 U.S.C. § 6801; 45 C.F.R. § 164.410.
[ ] Check for misleading security statements.
- Authority. 15 U.S.C. § 45.

Phase 7. Subprocessors and audit

[ ] Obtain the subprocessor list.
[ ] Obtain notice of additions, and a termination right where an objection cannot be accommodated.
- Why. Objection rights reduce to notice plus termination because a vendor cannot manage per-customer infrastructure.
[ ] Confirm model providers are named where machine learning is involved.
[ ] Set realistic audit expectations.
- Why. Certification reports plus a questionnaire is the middle; on-site audit is reserved for regulated customers or for cause.
[ ] Confirm where data actually resides, including backups and support access.

Phase 8. Term and renewal

[ ] [Gate] Cap the price increase.
- Why. Tied to a published index or a fixed percentage. The largest long-term exposure in most portfolios, and it costs the vendor nothing at signature.
[ ] [Gate] Calendar the non-renewal notice date on the day of signature.
- Trap. More value is lost to a missed notice window than to any negotiated term.
[ ] Consider term length honestly.
- Why. A multi-year commitment buys a discount and removes the annual exit. For a system of record that trade deserves more thought than it usually gets.
[ ] Ask for termination for convenience. Rarely granted and worth asking for.
[ ] Obtain termination for cause with a cure period, plus the chronic-failure right.
[ ] Address the unilateral amendment clause.
- Why. Whether the vendor may change terms by posting, whether notice is required, and whether the only remedy is termination. It quietly rewrites everything else between renewals.

Phase 9. Exit

[ ] [Gate] Specify the data export.
- Why. Format, completeness including custom fields, attachments, and audit history, mechanism, window, and cost.
- Trap. "Standard export functionality," which is not a commitment.
[ ] [Gate] Obtain transition assistance.
- Why. Continued access at the existing rate for a defined period after termination, and cooperation with a successor vendor.
[ ] Require deletion certification after the export window.
[ ] Do not rely on source code escrow for a hosted service.
- Why. Code without infrastructure and operational knowledge is not usable. A hosted continuity arrangement or a right to run the software on the customer's infrastructure is worth more.
[ ] Address vendor insolvency, including treatment of the agreement in bankruptcy.
[ ] Test the exit before signature.
- Why. Ask the vendor to describe in writing exactly what a departing customer receives. The answer is informative either way.

Phase 10. Statements of work and custom development

[ ] [Gate] Take a written assignment.
- Authority. 17 U.S.C. § 101; 17 U.S.C. § 201.
- Why. The developer owns what it writes absent a written assignment; the work made for hire categories are narrow and a services contract is not automatically within them.
[ ] Use present-tense assignment language.
[ ] Take a license back to vendor background technology needed to use the deliverable.
[ ] Define deliverables and acceptance criteria, with a process and a remedy for failure.
[ ] Read the statement of work's own terms.
- Trap. It frequently incorporates weaker terms than the master agreement or purports to supersede it.
[ ] Address open source in deliverables.
- Why. A component's license conditions travel with the code regardless of the vendor's warranties.
[ ] Define the developer's reuse carve-out rather than leaving it open.

Phase 11. The AI provisions

[ ] Determine whether customer content may be used to train.
[ ] Read the output ownership clause, including the caveat that similar outputs may be generated for others.
[ ] Understand the authorship gap.
- Authority. 17 U.S.C. § 102.
- Why. Purely machine-generated material may not be copyrightable, so an assignment conveys whatever rights exist rather than a guaranteed copyright.
[ ] Read the output indemnity's conditions.
- Trap. Conditioned on using the service as directed and not disabling filters. The conditions carry the clause.
[ ] Reconcile input restrictions with the customer's own compliance obligations.
[ ] Address human review of inputs and outputs, and whether it can be disabled.
[ ] Obtain notice of material model changes.
- Why. A customer that validated behavior against one model has no commitment that it persists.

Phase 12. Adjacent structures

[ ] On-premises perpetual license. A copy is delivered, so 17 U.S.C. § 117 applies and the exit risk is obsolescence rather than loss of access.
[ ] Term license, self-hosted. The exit analysis of a subscription plus the operational burden of on-premises.
[ ] Managed services. A three-party arrangement where flow-down governs, and where the manager's obligations are frequently weaker than the underlying license requires.
[ ] Reseller and marketplace purchases. The vendor's terms incorporated by reference while the reseller disclaims everything.
[ ] API and developer terms. Rate limits, permitted uses, retention restrictions, and unilateral change rights, governing integrations the business depends on.
[ ] Hardware with embedded software. Read the software license before evaluating the hardware warranty.

Phase 13. Portfolio management

[ ] Maintain a register.
- Why. Vendor, service, spend, term, renewal date, notice date, price cap, exit terms, and data sensitivity.
[ ] Write a position sheet with preferred, acceptable, and walk-away language for the twelve terms that matter.
[ ] Set a spend threshold for full review, with the triage below it.
[ ] Standardize the security addendum and data processing agreement so the customer proposes rather than reacts.
[ ] Keep a concession log.
- Why. The next negotiation with the same vendor starts from the last one, and vendors remember even when customers do not.
[ ] Review at the second renewal.
[ ] Report the portfolio annually.
- Why. Total spend, agreements without price caps, agreements without exit terms, and agreements renewing in the next two quarters. One page, and it is what gets exit terms taken seriously next time.

Phase 14. The twenty-minute triage

[ ] One: the renewal clause. Notice period, auto-renewal, price cap. Two minutes.
[ ] Two: the exit clause. Export format and completeness, transition assistance, deletion. Three minutes.
[ ] Three: the data clauses. Ownership, usage and aggregated rights, model training. Four minutes.
[ ] Four: the liability cap and carve-outs. Three minutes.
[ ] Five: user and affiliate definitions against how the business works. Three minutes.
[ ] Six: service level exclusions and whether credits are exclusive. Two minutes.
[ ] Seven: incident notification deadline. One minute.
[ ] Eight: the unilateral amendment clause. Two minutes.
[ ] Record the eight answers in the register.
- Why. Run across a portfolio, this identifies renewals needing attention a quarter in advance and surfaces the agreements nobody has read since signature.

Phase 15. What the vendor is protecting

Reading the form from the vendor's side explains most of the terms customers find objectionable, and it tells a negotiator where to spend.

[ ] Accept that uniformity is non-negotiable.
- Why. A vendor operating one multi-tenant system cannot maintain per-customer infrastructure, subprocessors, or security configurations. Terms requiring it will be refused regardless of leverage.
[ ] Understand the liability position.
- Why. Subscription economics assume liability bounded by fees. Movement is a function of deal size and insurability, which is why a security super-cap is obtainable and unlimited liability is not.
[ ] Understand the competitive restrictions.
- Why. Benchmarking, competitive-use, and reverse engineering prohibitions exist because the vendor's most dangerous customer is a future competitor.
[ ] Understand the improvement loop.
- Why. Usage data, aggregated analytics, and feedback feed product development. The negotiation is about scope and identifiability, not about whether the vendor gets anything.
[ ] Understand operational commitments.
- Why. Support targets, uptime, and notification deadlines have staffing consequences. A vendor agreeing to a two-hour notification obligation must staff for it, which is exactly what makes it a real trade.
[ ] [Gate] Sort the asks into what costs the vendor money and what does not.
- Why. Price caps, exit assistance, indemnity carve-outs, and data-use limits are frequently obtainable. Per-customer infrastructure, unlimited liability, and unilateral audit over a multi-tenant environment generally are not.

Phase 16. Multi-tenant realities to accept

[ ] There is one version. Everyone upgrades together; a customer cannot stay on an old release or require testing before changes reach production. Advance notice and a sandbox are the realistic asks.
[ ] Configuration is not customization. What the customer can change is what the vendor built configurable.
[ ] Integrations depend on vendor interfaces, and stability is rarely committed. Ask for a deprecation notice period.
[ ] Performance is shared, and degradation frequently does not count as downtime.
[ ] Data location may not be what the customer assumes, given multi-region deployments, backup replication, and support access.
[ ] Subprocessor changes are a fact of operation.
[ ] Conclude accordingly.
- Why. The customer cannot control the service. It can control whether it can leave, which is why Phase 9 carries the weight the operational phases cannot.

Phase 17. Where the money goes

[ ] Renewal pricing. The largest long-term exposure in most portfolios.
[ ] Overage and true-up, driven by the user and affiliate definitions.
[ ] Modules and add-ons priced on the assumption that the customer is already committed.
[ ] Professional services on time and materials, exceeding the estimate, under weaker terms.
[ ] Exit, paid later at the vendor's convenience if not negotiated at the start.
[ ] [Gate] Allocate the negotiation budget accordingly.
- Why. A customer with limited leverage should spend it on the renewal cap and the exit provisions rather than on the liability cap. The liability cap matters if something goes wrong; the renewal and exit terms matter every year regardless.

Phase 18. Renewals, and the leverage that returns

Leverage disappears at signature and comes back once a year, briefly, and only for a customer that prepared.

[ ] Start twelve months out for anything material.
- Trap. A renewal negotiated in the last thirty days is not a negotiation.
[ ] [Gate] Establish whether the alternative is real.
- Why. A documented evaluation of a competitor, with an implementation estimate and a migration plan, converts a request into a negotiation. Without it, the vendor prices to the customer's inability to move.
[ ] Test the export before the conversation.
- Why. Run an actual export and see what comes out. Discovering the format is unusable in month twelve is better than in month thirty-six.
[ ] Bring the usage data.
- Why. Seats provisioned versus active, modules purchased versus used, and consumption against entitlement. Vendors expect a price discussion and are unprepared for a utilization discussion.
[ ] Ask for the terms that were refused at signature.
- Why. A vendor facing renewal risk will move on exit assistance and price caps it declined three years earlier.
[ ] Reassess the term trade.
- Why. A multi-year commitment buys a discount and removes two annual exits.
[ ] Fix what was learned.
- Why. The affiliate definition, the notification deadline, and the export specification. Renewal is the only opportunity most agreements ever get.

Phase 19. Talking to the business

[ ] Explain that a subscription is a dependency, not a purchase.
- Why. No copy, no continuation on termination, no transferable rights. Everything about the exit terms follows, and a business owner who understands it stops treating them as lawyerly caution.
[ ] Explain that leverage exists only before signature.
- Why. After signature the vendor's incentive to accommodate is a function of renewal risk, which is a function of whether the customer can actually leave.
[ ] Explain that the negotiation has a budget.
- Why. Six asks get engaged; sixty get a form response. The business should help choose which six, because at least two are commercial rather than legal.
[ ] Ask the business five questions before drafting.
- Why. How dependent will we become; what data goes in; how many people and which entities; what would migration cost; and what is the realistic alternative if this vendor becomes unacceptable.
[ ] Report back after signature, in the register.
- Why. What was obtained, what was conceded, when the notice date falls, and what the exit actually looks like. The people who negotiated it will not be the ones facing the renewal.

Phase 20. Failure modes, collected

[ ] Redlining everything. A markup touching a hundred clauses gets a form response and consumes the review budget.
[ ] Negotiating the license grant rather than the exit.
[ ] Ignoring the affiliate definition.
[ ] A named-user agreement for occasional broad usage.
[ ] Unlimited scheduled maintenance accepted as an uptime commitment.
[ ] Negotiating larger credits instead of a termination right.
[ ] Renewal pricing left uncapped.
[ ] The non-renewal notice window missed.
[ ] A multi-year term committed to without noticing it removed the annual exit.
[ ] "Standard export functionality" accepted as an exit commitment.
[ ] No transition assistance, so migration happens under time pressure at the vendor's convenience.
[ ] Indemnity inside the liability cap.
[ ] No security-incident super-cap.
[ ] Incident notification measured in days.
[ ] The unilateral amendment clause ignored.
[ ] The statement of work assumed to be covered by the master agreement.
[ ] Payment assumed to convey ownership of custom development.
- Authority. 17 U.S.C. § 201.
[ ] Source code escrow relied on for a hosted service.
[ ] The exit never tested before signature.
[ ] Model training permitted without evaluating the content's confidentiality classification.
[ ] Output indemnity conditions never checked against the customer's actual workflow.
[ ] API and developer terms never read before building an integration the business depends on.

Phase 21. Building the standing playbook

[ ] Write the position sheet.
- Why. For each of the twelve terms that matter, the preferred language, the acceptable compromise, and the walk-away. It turns most negotiations into lookups and lets business people resolve routine ones without counsel.
[ ] Set a spend threshold for full review, with the twenty-minute triage below it.
[ ] Standardize the security addendum and data processing agreement.
- Why. So the customer proposes rather than reacts.
[ ] Keep the concession log.
[ ] Calendar every notice date on the day of signature, in a system that alerts the business owner and not only the lawyer.
[ ] Review at the second renewal.
[ ] Report the portfolio annually.
- Why. Total spend, agreements without price caps, agreements without exit terms, and agreements renewing in the next two quarters. One page, and it is the document that gets exit terms taken seriously the next time.
[ ] Train the people who sign.
- Why. Most technology agreements are executed by business owners under a delegated authority, and the single highest-return intervention is teaching them the eight triage questions.

Phase 22. Insurance, escalation, and the relationship

[ ] Check the customer's own coverage.
- Why. Technology errors and omissions, cyber, and general liability each respond differently to a vendor failure, and a customer relying on a vendor's cap should know what its own policy covers.
[ ] Ask for the vendor's insurance certificates.
- Why. Coverage amounts and types, renewed annually, with the customer named as an additional insured where the size of the deal supports asking for it.
[ ] Build an escalation path into the agreement.
- Why. Named contacts at defined levels, with response times, for issues that are not service level breaches but are business problems. This is the mechanism that resolves most disputes before they become legal ones.
[ ] Agree a governance cadence for material relationships.
- Why. Quarterly business reviews with usage, performance, and roadmap. A customer that meets its vendor regularly gets earlier notice of changes and better treatment at renewal.
[ ] Keep the relationship separate from the paper.
- Why. A good relationship does not fix bad exit terms, and good exit terms do not require an adversarial relationship. Customers conflate these constantly, treating a cooperative vendor as a reason not to negotiate.
[ ] Document what was promised outside the agreement.
- Trap. Roadmap commitments, implementation assurances, and performance representations made in the sales process, none of which survive an integration clause unless they are written into the agreement or into a side letter signed by someone with authority to bind the vendor.

Outcomes. The manufacturer's six asks produced a price cap in the first round, a four-month transition period with a full export specification, a chronic-failure termination right traded for the credits, a security super-cap at three times annual fees after two rounds, a model training prohibition with an aggregated-telemetry carve-out, and affiliate inclusion for one subsidiary at no charge. Five weeks, three rounds, eleven clauses redlined. The company at its fourth renewal had none of these: "standard export functionality" produced records without custom fields or attachments, migration would have required nine months of parallel operation the vendor had no obligation to support, and it renewed on the vendor's terms because leaving cost more than staying.

Key Authorities at a Glance

| Authority | Proposition | Phase | |---|---|---| | 17 U.S.C. § 101 | Work made for hire definitions | 10 | | 17 U.S.C. § 102 | Subject matter; authorship | 11 | | 17 U.S.C. § 106 | Exclusive rights | 2 | | 17 U.S.C. § 109 | First sale; inapplicable without a copy | 2 | | 17 U.S.C. § 117 | Essential step and archival copies | 2, 12 | | 17 U.S.C. § 201 | Ownership; written assignment | 10 | | 17 U.S.C. § 1201 | Circumvention; interoperability | 2 | | 18 U.S.C. § 1030 | Computer access | 2 | | 18 U.S.C. § 1836 | Trade secret civil action | 4 | | 18 U.S.C. § 1839 | Trade secret definition | 4 | | 15 U.S.C. § 45 | Unfair or deceptive practices | 6 | | 15 U.S.C. § 6801 | Financial safeguards | 6 | | Cal. Civ. Code § 1798.100 | Notice and purpose limitation | 6 | | Cal. Civ. Code § 1798.105 | Deletion capability | 6 | | Cal. Civ. Code § 1798.140 | Service provider definitions | 6 | | 45 C.F.R. § 164.410 | Business associate breach notice | 6 | | Fed. R. Civ. P. 65 | Injunctive relief | 5 |

The five things people get wrong

One: they negotiate the license grant. It is the section customers understand and the one that matters least. A subscription conveys access, not a copy, and 17 U.S.C. § 117 does not apply — which is why the exit terms carry the weight the grant does not.

Two: they leave renewal pricing uncapped. Increases of twenty or thirty percent are ordinary, migration is expensive enough that the customer pays, and a cap costs the vendor nothing at signature.

Three: they accept "standard export functionality" as an exit commitment. It produces records without custom fields, attachments, or audit history, and the discovery comes at the moment the customer most needs to leave.

Four: they ignore the affiliate definition. Subsidiaries get connected, the annual reconciliation produces an unbudgeted invoice, and the invoice is correct under the agreement as written.

Five: they leave indemnity inside the liability cap. An intellectual property indemnity capped at twelve months of fees does not answer a patent claim, and breach exposure routinely exceeds the same cap — which is why a security super-cap is the standard compromise. See Negotiating a Technology Agreement.