inxo — Terms of Service

Version 2026-10-01. For an organization that accepted an earlier version, this version takes effect on 2026-11-05, or when that organization accepts it, if sooner (§9).

These Terms of Service (“Terms”) govern use of the inxo service (“inxo”, the “Service”) provided by No Compromise AI, LLC (“we”, “us”). “You” and “Customer” mean the organization on whose behalf the Service is used.

---

1. Agreement, and the authority to make it

By clicking to accept, the individual accepting represents that they are authorized to bind the organization they name to these Terms and to the Data Processing Addendum. This representation is asked for explicitly because, unlike the authority to connect a mailbox — which cannot be faked, since the connection cannot complete without genuine Exchange administrative authority in your tenant — contracting authority leaves no comparable trace.

If you do not have that authority, do not accept. An acceptance is recorded against your account, append-only, with the document, its version, a content hash of the exact text accepted, the time, and the accepting individual.

2. What the Service does

inxo connects to a Microsoft 365 mailbox you control, reads and threads the mail that arrives there, and classifies each message it sees. For the mail that calls for a reply — the customer correspondence your team would otherwise have answered by hand — it drafts one. Not every message gets a drafted reply: spam, bounces and similar non-correspondence get none, and some classes of message get a short acknowledgement rather than a substantive answer. The AI disclosure describes which is which, because the distinction matters to what you can expect from the Service.

Drafting a message is not sending it. It does not decide what to say to your correspondents on its own account: it proposes text, and a person at your organization reviews it before it goes out, unless you have separately enabled automated sending under §4.

3. You are the author and the sender — this is the substance of the arrangement

Every message sent through the Service is your message. You are the author and the sender of record. We are not.

This is not a disclaimer bolted onto a product that behaves otherwise; it is what the product is built to make true:

• A draft is a proposal until a human adopts it. Human approval is the default and cannot be turned off by accident. Approving is the act of authorship — the person approving is taking editorial responsibility for the words, and the approval is recorded.
• You may edit any draft before approving it, and editing changes the body only. The recipient is derived by the Service from the thread being replied to and is never supplied by a client, so an edit cannot redirect a message to somebody else.
• The Service works to keep certain things out of replies — commitments to a fix, promised timelines, internal identifiers, and invented facts. These are not all enforced the same way, and the AI disclosure says which is which: some are structurally impossible, some are blocked by a check on the finished text, some are checked and recorded without blocking, and some are steered and measured with your reviewer as the actual control.
• Whether replies may mention your own products is your setting, not ours. Replies are transactional by default — they answer what was asked — and every workspace starts that way. You may turn on the ability for drafts to mention your own products, plans or pricing. Turning it on relaxes no other instruction: the model is still instructed not to invent an offer, a price, a discount, or a link it was not given — though that, like the default itself, is an instruction rather than a detector, and the AI disclosure says so in terms. We give you the default and the switch rather than a content filter because a filter could tell you a sentence reads as promotional but not whether sending it was lawful — that turns on your recipient’s relationship with you, on any exemption you rely on, and on the identification and unsubscribe requirements that attach if the message is commercial. It matters commercially as well as legally: a call to action can turn a transactional reply into a commercial electronic message under CAN-SPAM or CASL, and CASL and the EU/UK rules are stricter than the US ones. That liability is yours, not ours — which is why the decision is yours too.
• And one limit you need before you decide. Where a message is commercial, the rules generally also require it to identify you and offer a way to opt out. inxo cannot guarantee either appears — we attach nothing to a reply after the model writes it, deliberately, since a footer would land on ordinary transactional replies too. Wording you put in your workspace voice is an instruction, not a guarantee. If your replies need to carry that wording to be lawful where your recipients are, do not turn this on unless a person reads every reply. The AI disclosure says the same thing at more length.

You warrant that the correspondence you direct through the Service is lawful — which, where it is commercial, includes any consent, identification and unsubscribe requirements that apply to it. Ordinary use is transactional: replies to people who wrote to you. Turning on product mentions is permitted by these Terms and is not itself a breach of this warranty. What you must not do is send commercial correspondence to a recipient for whom you have no lawful basis to do so — consent, an existing business relationship, or another exemption you are entitled to rely on.

You will indemnify us against third-party claims and regulatory demands arising from a breach of that warranty, including a claim that a message sent through the Service was an unlawful commercial electronic message.

What that does not cover. This indemnity does not cover a claim to the extent it results from our own failure of a control this agreement or the AI disclosure describes as structural, rather than from your configuration, your content, your correspondence, or your judgment — specifically: our failure to derive a message’s recipient from the conversation being replied to; our failure to confine what the model produces to the body text of the reply; our sending a message without the required AI marking instead of failing to send it; our failure to run the final check on the exact text of a message before it is sent, or to act on what that check found where the check is one that blocks; and a failure of the database-level isolation that keeps one customer’s data from being returned by a query for another’s. It also does not cover a claim to the extent it results from our model inventing an offer, a price, a discount, or a link it was not given — we instruct it not to, we tell you plainly that the instruction is not a detector and can be broken, and we are not going to describe a failure in those terms and then charge it to you. A message affected both by one of these failures and by your own content or configuration is apportioned between us accordingly, not treated as entirely ours or entirely yours.

This mirrors the indemnity in the authorship election, deliberately, and that election carries the same exclusions. The two documents describe the same kinds of failure and allocate them the same way — it would not be honest for the higher-risk automated-sending mode to carry a protection that ordinary use does not, nor the other way round.

4. Automated sending is a separate decision, and you have not made it here

The Service can be configured so that a message is sent without a person approving it first. That mode is off by default, must be enabled per workspace, and is materially higher risk: the human review that makes §3 work is exactly what it removes.

Accepting these Terms does not enable it and does not elect it. Enabling automated sending requires a separate, explicit authorship election taken at the time you enable it, which states in its own words what you are accepting responsibility for. We keep it separate deliberately — folding a higher-risk liability election into routine signup boilerplate would both under-disclose it and destroy its value as evidence that you chose it on purpose. That election also contains an indemnity: enabling automated sending means agreeing to defend and indemnify us against claims arising from the messages it sends, subject to the carve-outs stated there for our own structural failures.

5. Connecting a mailbox

You warrant that you are authorized to connect each mailbox you connect, and that the administrative consent granted for it is properly given within your organization. Where you operate a workspace on behalf of a client — an agency or managed-service arrangement — you additionally warrant that you hold that client’s authorization to instruct us on their behalf, and you will indemnify us against claims arising from a failure of that authority.

A connection reaches the active state only after the Service verifies your effective permissions live against Microsoft, and only where the processing instruction evidencing the client’s own controller relationship exists. An account-level acceptance does not stand in for a client’s instruction where the client is a different organization.

6. Intellectual property

You own your inputs and your outputs. Mail that arrives in your mailbox, the text you write, and the drafts generated for you are yours. We take only the narrow licence needed to operate and secure the Service for you — to store, process, transmit and display your content for the purpose of providing it — and for no other purpose.

We do not train models on your data. We do not use your content, your drafts, or your correspondents’ mail to train, fine-tune or improve any model, and we will not — regardless of what any model provider’s terms say. We will not send your mail to a model provider until that provider’s no-training commitment is executed and in force for the specific model endpoint used. That is not a promise we ask you to take on faith: a fail-closed readiness check keyed to the specific provider, model and region blocks drafting and classification for a workspace until the provider’s terms are in force for its target — there is no path by which your mail reaches an endpoint whose no-training terms are not yet executed.

Infringement. We will defend you against a third-party claim that the Service, used as provided and in accordance with these Terms, infringes that party’s U.S. patent, copyright or trademark, or misappropriates its trade secret, and will pay the damages and costs finally awarded against you (or agreed in settlement with our written consent). This does not cover a claim arising from your content, a modification of the Service you made or directed, your breach of these Terms, or a combination of the Service with something we did not provide, to the extent the claim arises from that combination. If we reasonably believe such a claim is likely, we may, at our option and expense: procure the right for you to keep using the affected part of the Service; replace or modify it so it is no longer claimed to infringe, without materially reducing its functionality; or terminate the affected subscription and refund the fees you prepaid for the period after termination. This is your exclusive remedy, and our sole liability, for a claim of this kind.

What that defence is not. It protects you against a claim that our platform — the connecting, threading, classification and drafting technology itself — infringes someone’s rights. It is not an indemnity for the content of a particular message: because you own the drafts the Service generates for you and you are the author and sender of record for what is actually sent (§3), a claim that the wording of a specific draft or sent message infringes a third party’s copyright, trademark or other right is a claim arising from your content, and the exclusion above for “a claim arising from your content” applies to it. We say this plainly rather than leave it to be inferred: we do not offer a generated-output copyright indemnity, because doing so would sit uneasily beside a product built on the premise that you are the author of what you send, not us. If protection against that specific risk matters to your business, say so before you rely on the Service for correspondence where it does.

7. Acceptable use

Do not use the Service to send unlawful, deceptive or harassing mail, to impersonate another person or organization, or in a way that violates Microsoft’s terms for the mailbox you have connected. Do not use it to send unsolicited commercial mail in bulk — by which we mean commercial messages sent to recipients who did not write to you and for whom you have no lawful basis to send them. A reply to a correspondent who contacted you is not that, even where you have turned on product mentions under §3 and the reply mentions your own products: §3 governs that case, and using the setting as it is described there is not a breach of this section. We may throttle, suspend or quarantine an account or workspace where we detect abuse, and you may appeal such a decision.

A throttle, suspension or quarantine under this section is an operational safety measure, not a termination of these Terms, and applying one does not by itself waive, or substitute for, our separate right to terminate for cause under §10 where the underlying conduct is also a material breach.

Sanctions and export control. Each party represents and warrants that it is not identified on the U.S. Treasury’s Specially Designated Nationals and Blocked Persons List, or on any other applicable list of sanctioned, debarred or denied parties maintained by the United States, the United Nations, the European Union or the United Kingdom, and is not organized in, headquartered in, or ordinarily resident in a country or region subject to a comprehensive U.S. embargo. You additionally represent that, to your knowledge, this is true of any individual who will use the Service on your behalf, and that no person or entity described above owns fifty percent (50%) or more of, or otherwise controls, your organization. Each party will comply with applicable export control and economic sanctions law in connection with these Terms, and you will not use the Service in a way that would cause us to violate that law. This representation is made continuously throughout the term, not only at acceptance.

How we check this. Before your account can send mail, and periodically after, we screen the identity information you give us against sanctions and embargoed-jurisdiction lists, and a match holds or suspends the account rather than releasing it. The Privacy Policy describes how that screening works, what it does and does not cover, and how to contest a match; nothing here narrows that description, and nothing here claims we screen against every list the representation above names. A hit under that screening — or our own reasonable determination that continuing to provide the Service to you would violate applicable law — is cause for immediate suspension or termination under §10, without the cure period that section otherwise provides, because continuing to process your mail in that circumstance is not something we are free to cure our way past.

8. Fees

Fees, billing periods and plan entitlements are those stated at the time of purchase. Payment is handled by our payment processor; we do not receive or store your card details.

Auto-renewal. A subscription renews automatically at the end of its billing interval — monthly or annual, whichever you selected — at the then-current price for your plan, unless you cancel before the renewal date. You may cancel at any time in the Console; cancellation takes effect at the end of the billing interval already paid for, and we do not refund the unused portion of a period you have already paid for, except where law requires it or as §10 provides on a termination for cause or for a sustained force-majeure event. We will make reasonable efforts to email you in advance of an annual renewal and of any price increase that will apply to it.

9. Changes to these Terms, the AI disclosure, the Data Processing Addendum, the Privacy Policy and the authorship election

We may revise these Terms, the AI disclosure, or the Data Processing Addendum. Each is published separately, as a new version with a new content hash, and your acceptance records the hash of the text you actually agreed to, so which version you agreed to always stays answerable. The same publication, notice and early-acceptance mechanism applies to the Privacy Policy and to the authorship election under §4, even though neither is part of these Terms.

Notice, and when a revision takes effect. On the day we publish a revision to any of these documents, we email every administrator on your account and post a notice in the Console describing what is changing, with an effective date at least 30 days later; §12 describes how that notice is given and says exactly when the notice period runs from. Until the effective date, the version your organization most recently accepted remains the agreement in force for that document and continues to authorize what it authorized. A revision does not reach backward.

Accepting early. An administrator may accept a revision at any time before its effective date. Accepting it early means the revised version governs, for your organization, from the moment of that acceptance — not from the later effective date — while anyone who has not yet accepted continues under the version before it until the effective date arrives on its own. Accepting a revised Data Processing Addendum early additionally rolls your account’s processing instruction over to the revised text at that moment, as the Addendum’s own §13 describes.

Which capabilities stop once a revision takes effect and you have not accepted it — stated exactly, because we would rather be precise than reassuring. A stale authorship election stops automated sending, until the current election is accepted again. A stale version of these Terms or of the AI disclosure keeps a workspace from switching product mentions on, until the current version of both is accepted — the warranty in these Terms is the control for that setting, so the setting is not available to an account that has not accepted it. A stale Privacy Policy does not stop anything for an existing account — a new account accepts the current version when it signs up, but it is a disclosure, not a condition of continued use — though you receive the same notice before a revision of it takes effect. And a stale Data Processing Addendum does something larger than disable a feature, described in its own paragraph below. None of the rest reaches drafting, human-approved sending, editing your voice, or switching product mentions off — those continue regardless, because cutting off a working inbox over a revised text would be worse for you than the stale text is. That is a deliberate limit, not an oversight, and we say so here rather than describe a stoppage that does not happen.

The Data Processing Addendum is different, and it is not us being cautious — it is how the processing gate works. If a revision to it is not accepted by its effective date, processing pauses for the controller scope whose processing instruction has lapsed — your account’s own mailboxes and knowledge base, or, where a client workspace’s own processing instruction under §5 was not renewed, that workspace alone (scopes are independent of each other). Reading, threading, classifying, drafting, sending, and knowledge-base filing, retrieval and re-indexing all stop for that scope — the knowledge-base half of this reaches a workspace even before any mailbox is connected to it, because filing and retrieval run on the same instruction regardless. The pause does not delete anything and does not disconnect a mailbox by itself, and it ends the moment an administrator accepts — or, for a client workspace, the moment its own processing instruction is renewed. The reason is the same in every case: we are only allowed to process on a recorded instruction naming the current Data Processing Addendum, and a revision makes the recorded one no longer current once it takes effect. That is the same gate that let processing start in the first place, working in the direction it was built to work in. Fees continue to accrue during a pause, and we do not refund or credit the paused period — a pause is a consequence of the instruction lapsing, not a service failure, and §10’s refund rules do not treat it as one.

If the pause lasts 90 consecutive days. Where processing for a controller scope has been paused under this section for 90 consecutive days, we may terminate — these Terms in their entirety, or, where the paused scope is a single client workspace, these Terms as to that workspace only — on 30 days’ further written notice. The Data Processing Addendum’s own deletion and return provisions (its §7) then apply to what is terminated, the same as on any other termination. §10 lists this among our termination rights and states the refund that follows.

Termination for a material revision of the Data Processing Addendum. Where a revision to the Data Processing Addendum is a material one — the Addendum’s own §13 states the test, directly — you do not have to wait for the pause above, accept the revision, or let it lapse. At any time before you accept it, you may terminate without penalty by written notice to legal@inxo.io, in either of two forms: terminating these Terms in their entirety, or, where you operate a workspace on behalf of a client under §5, terminating these Terms as to that one client workspace only, which disconnects that workspace’s connections without affecting any other. §10 states the refund that follows and lists this termination among those it covers.

This revision. This is the revision that first states the termination right above. The version of these Terms you are revising from did not grant it. We are extending it to this revision anyway: if you consider this revision of the Data Processing Addendum material, you may exercise the right above with respect to it, on the same terms, even though the right is being created by the very revision you would be responding to.

Changes to our sub-processor register follow a different and deliberately gentler rule, described in the Data Processing Addendum: we notify you in advance and you may object, rather than blocking your mail over a routine register update.

10. Confidentiality, warranties, liability and term

Confidentiality. Each party may receive information from the other that is marked confidential or that a reasonable person would understand to be confidential given its nature and the circumstances of disclosure (“Confidential Information”) — for us, the non-public parts of our systems, security practices and pricing; for you, your non-public business information other than the personal data governed by the Data Processing Addendum, which that Addendum protects instead. Confidential Information does not include information that is or becomes public through no fault of the receiving party, was already known to it without a confidentiality obligation, is developed independently without reference to the disclosing party’s information, or is rightfully received from a third party free of restriction. The receiving party will use Confidential Information only to perform its obligations or exercise its rights under these Terms, will protect it with at least the care it uses for its own confidential information of similar sensitivity and never less than reasonable care, and will disclose it only to personnel, advisors and subcontractors who need it for that purpose and are bound to obligations at least as protective as these. A party may disclose Confidential Information where compelled by law, giving the other party notice first if the law allows it. This obligation survives for three years after disclosure, except that a trade secret remains protected for as long as it remains one under applicable law.

Warranties. The Service is provided as described in the documentation. Drafts are proposals; you are responsible for what you send — we do not warrant that a draft is accurate, complete, or fit for a particular purpose. Nor do we warrant that a draft, or a message sent from it, is free of a third party’s intellectual property rights — §6 explains why, and what protection we do and do not offer for that risk. EXCEPT AS EXPRESSLY STATED IN THESE TERMS, THE SERVICE IS PROVIDED “AS IS,” AND WE DISCLAIM ALL OTHER WARRANTIES, EXPRESS OR IMPLIED, INCLUDING MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT.

Limitation of liability. To the maximum extent permitted by law, and except as stated below, our aggregate liability arising out of or relating to these Terms or the Service is limited as stated in the order form or, absent one, to the fees you paid in the twelve months preceding the claim, and neither party is liable for indirect, incidental, special or consequential loss, or for lost profits or lost data.

What that cap does not limit. The cap above does not apply to, and does not limit, our liability for: (a) our gross negligence or willful misconduct; (b) our breach of the Confidentiality clause above; (c) either party’s indemnification obligations under these Terms or the authorship election; or (d) death or personal injury caused by our negligence, or any other liability that cannot be limited or excluded as a matter of law.

A separate, higher limit for a data-breach or confidentiality event. The ordinary cap above, on its own, could limit our liability for a failure of the safeguards that keep one customer’s data from another’s to a small fraction of what such a failure could actually cost you. So our aggregate liability arising from a Data-Breach Event — a breach of our security obligations under the Data Processing Addendum, a failure of the database-level tenant isolation it describes, or an unauthorized disclosure of the personal data or Confidential Information we hold for you — is instead the greater of (i) five times the fees you paid us in the twelve months preceding the event (annualized, if you have been a customer for less than twelve months), or (ii) $150,000, and in no event more than $1,000,000 in the aggregate for all of your claims arising from the same Data-Breach Event. This higher limit is itself subject to the carve-outs in the paragraph above, which remain uncapped.

Force majeure. Neither party is liable for a failure or delay in performing an obligation under these Terms — other than an obligation to pay money — to the extent the failure or delay is caused by an event beyond that party’s reasonable control that the party could not have avoided or mitigated through commercially reasonable diligence, including an act of God, war, terrorism, civil unrest, a government order or change in law, a labor dispute, an epidemic, pandemic or other public health emergency, or a failure or outage of the public internet or of a third-party platform, network or cloud infrastructure the Service depends on, including Microsoft 365 (a “Force Majeure Event”). The affected party will give the other prompt notice describing the event and its expected impact, and will resume performance as soon as reasonably possible once it ends.

What force majeure does not excuse. A Force Majeure Event does not excuse your obligation to pay fees already due — a delay in the Service is not a delay in what is owed for it, and your remedy for a sustained outage is the paragraph below, not withholding payment. It does not excuse either party’s confidentiality obligations. And it does not excuse the structural controls §§3 and 6 describe as such — deriving a message’s recipient from the conversation being replied to, confining what the model produces to the reply body, applying the required AI marking, running the final check on outbound text and acting on what it finds, and the database-level tenant isolation the Data Processing Addendum describes. Those controls exist specifically so that their failure is ours to answer for regardless of cause, and a Force Majeure Event does not change that allocation. If the actual cause of a Data-Breach Event was genuinely outside our security obligations — a failure our contracted safeguards could not have prevented — that goes to whether we breached those obligations in the first place, decided on the facts; it is not a separate defence once a breach is found.

A Force Majeure Event lasting more than 60 days. If a Force Majeure Event prevents a party from performing a material obligation for more than 60 consecutive days, the other party may terminate the affected subscription, or these Terms, on written notice, without further liability for the non-performance beyond fees already accrued to the date of termination — except that if we are the party prevented from performing, you are entitled to a pro-rata refund of prepaid fees for the period after the effective date of termination in which you did not receive the Service.

Term. These Terms remain in effect for as long as you maintain an account with an active subscription under §8, and continue to govern every workspace and mailbox connection under that account until these Terms are terminated as this section describes.

Termination for cause. Either party may terminate these Terms if the other party materially breaches these Terms (including the Data Processing Addendum) and does not cure that breach within 30 days after receiving written notice describing the breach in reasonable detail. Where the breach cannot be cured — including a breach of the Confidentiality clause above, an infringement of the other party’s intellectual property, or a sanctions or export-control representation under §7 that is or becomes false — the non-breaching party may terminate immediately on notice, without waiting for the cure period to run.

Termination for non-payment. If you fail to pay fees when due and do not cure that failure within 30 days after we give you written notice of it, we may terminate these Terms, effective immediately. We may also suspend your access to the Service for so long as fees remain unpaid, whether or not we ultimately terminate. Non-payment is treated as a breach under the preceding paragraph; this paragraph exists only to say so in terms, and does not create a different or shorter standard.

Termination for insolvency. Either party may terminate these Terms immediately upon written notice if the other party: becomes insolvent or admits in writing its inability to pay its debts as they become due; makes a general assignment for the benefit of creditors; has a receiver, trustee, administrator or similar official appointed over a substantial part of its assets; has a petition filed by or against it under any bankruptcy, insolvency or reorganization law of any jurisdiction that is not dismissed or stayed within 60 days; or ceases to operate in the ordinary course of business. No cure period applies to a termination under this paragraph.

Termination for a material revision of the Data Processing Addendum. You may terminate as §9 and the Addendum’s own §13 describe — these Terms in their entirety, or, where you operate on behalf of a client under §5, these Terms as to that one client workspace only — at any time before you accept a material revision to the Data Processing Addendum, by written notice to legal@inxo.io. No cure period applies: there is no breach to cure, only your election not to proceed under a revised text you have not agreed to, exercised on the schedule and in the form §9 and the Addendum state.

Termination after a sustained processing pause. Where processing for a controller scope has been paused under §9 for 90 consecutive days because a revision of the Data Processing Addendum was not accepted, we may terminate these Terms — in their entirety, or, where the paused scope is a single client workspace, as to that workspace only — on 30 days’ further written notice, as §9 describes. No cure period applies: the 90 days already given, and the 30 further days of notice, are the opportunity to cure.

Effect of termination. On termination of these Terms for any reason: your right to access and use the Service ends immediately; we will handle the content we hold for you as described in the deletion and return provisions of the Data Processing Addendum; and you remain responsible for paying fees accrued, and not yet paid, through the effective date of termination. Termination under this section is separate from, and does not require, the cancellation process described in §8 — it takes effect on the date stated in the termination notice, or, where a cure period applies and is not satisfied, upon that period’s expiration, rather than at the end of a billing interval. Where these Terms are terminated as to a single client workspace only (§9), these effects apply to that workspace alone: its connections are disconnected and its content is handled as the Data Processing Addendum describes, and the rest of your account continues under these Terms.

Refunds on termination. Except as this paragraph provides, fees already paid are not refunded, and fees that accrue while processing is paused under §9 are not refunded or credited for that reason even if these Terms are later terminated — a pause is a consequence of a lapsed instruction, not a service failure. Subject to that, a pro-rata refund of the prepaid fees attributable to the period after the effective date of termination, prorated to whatever is terminated, is due in each of the following cases: where these Terms are terminated for our uncured material breach; where they are terminated under the force-majeure paragraph above and we are the party that was prevented from performing; where you terminate for a material revision of the Data Processing Addendum under §9 — prorated to the single client workspace terminated, where that is the form of termination you choose; or where we terminate after a sustained processing pause under §9 — prorated the same way where the scope terminated is a single client workspace. For the force-majeure case, which party gives the notice does not change who was prevented from performing, and it is the latter that decides the refund. No refund is due where we terminate for your breach, your non-payment, your insolvency, or to comply with law under §7’s sanctions and export-control paragraph; where these Terms are terminated under the force-majeure paragraph and you are the party prevented from performing; or where you terminate for convenience — that is, for any reason other than the cases named above. In those cases §8’s ordinary no-refund rule continues to apply.

Disconnecting a mailbox. You may stop using the Service and disconnect a mailbox at any time. On disconnection, we stop syncing that mailbox and delete the content we hold for that connection as described in the Data Processing Addendum, and we give you a record of what was deleted. The copy in your own Microsoft 365 mailbox is yours and remains yours — see the Addendum for exactly what our erasure does and does not reach.

Disconnecting does not, by itself, revoke the permission your Exchange administrator granted us in your own tenant. That permission is a role assignment that lives in your Microsoft 365 tenant, not in ours — true of any third-party application connected this way, because of how Microsoft’s application-level Exchange role assignments work (the connection is app-only, granted to the inxo application itself rather than delegated through a signed-in person), not a gap specific to inxo, and we cannot revoke it from our side. If you are disconnecting because you are concerned about our access specifically — including in response to a security incident — you must separately have your Exchange administrator revoke the role assignment granted to the inxo application; we will tell you which role to look for, and confirm once we can see it is gone. Until that revocation happens, the underlying permission remains live in your tenant even though the Service has stopped using it.

11. Governing law and dispute resolution

Governing law. These Terms, the Data Processing Addendum, and any dispute arising out of or relating to them or to the Service (a “Dispute”) are governed by the laws of the State of Delaware, without regard to its conflict-of-laws rules. This is the law that governs the agreement between us. It is a different thing from the internal affairs doctrine, which governs disputes about No Compromise AI, LLC’s own internal governance — among its members and managers — under the Delaware Limited Liability Company Act. That doctrine applies to us because we are organized in Delaware, by operation of law, not because this clause says so, and it has nothing to do with our agreement with you; you are not a party to it, and nothing here purports to change it.

Agreement to arbitrate. Except for the claims carved out below, you and we each agree to resolve any Dispute exclusively through final, binding, individual arbitration rather than in court. This section is governed by the Federal Arbitration Act. YOU AND WE EACH WAIVE THE RIGHT TO A JURY TRIAL AND TO LITIGATE A DISPUTE IN COURT, EXCEPT AS THIS SECTION PROVIDES.

How arbitration works. Arbitration is administered by the American Arbitration Association (“AAA”) under its Commercial Arbitration Rules then in effect, before a single arbitrator — or, for a Dispute in which the amount claimed exceeds $1,000,000, a panel of three arbitrators as the AAA’s rules provide — seated in Wilmington, Delaware. Either party may elect to appear by videoconference or on written submissions to the extent the AAA’s rules allow. The arbitrator, not a court, decides threshold questions about the formation, validity, enforceability or scope of this arbitration agreement, except a challenge to the enforceability of the class-action waiver below, which only a court may decide.

Fees. Each party pays its own attorneys’ fees and costs, except as the arbitrator awards under applicable law or an indemnification obligation in these Terms. Administrative and arbitrator fees follow the AAA’s Commercial Fee Schedule, except that if your total claim is less than $10,000, we will pay the AAA’s administrative fee and the arbitrator’s fee, so that arbitrating costs you no more than filing the same claim in Delaware small-claims court would.

Mass filings. If 25 or more similar demands for arbitration against us are filed by, or with the involvement of, the same or coordinated counsel within a 30-day period (a “Mass Filing”), the following applies instead of the ordinary process, using the AAA’s mass-arbitration procedures then in effect or, absent an applicable AAA procedure, the process in this paragraph: the demands are organized into batches of no more than 50; batches proceed sequentially, one active batch at a time, with the rest of the Mass Filing stayed pending resolution of the active batch; fees for a Mass Filing follow the AAA’s mass-arbitration fee schedule then in effect; and, informed by the outcome of each resolved batch, the parties will meet and confer in good faith about resolving the demands that remain. A Mass Filing administered this way is not a class, collective or representative proceeding, and does not adjudicate the rights of anyone who is not a named party to it.

What is not arbitrated. Either party may instead bring an individual claim in small-claims court — yours, in the county where your organization has its principal place of business, or ours, in New Castle County, Delaware — if the claim qualifies for that court. Either party may also seek injunctive or other equitable relief, in the courts identified below, to prevent actual or threatened infringement, misappropriation or violation of its intellectual property rights or of the Confidentiality clause of these Terms, pending the outcome of arbitration. For anything else not subject to arbitration under this section — including an action to compel arbitration, or to confirm, modify or vacate an award — you and we each consent to the exclusive jurisdiction and venue of the Court of Chancery of the State of Delaware or, if that court lacks subject-matter jurisdiction, the Superior Court of the State of Delaware, in each case sitting in New Castle County, and the U.S. District Court for the District of Delaware for any matter it may hear, and waive any objection to that venue, including forum non conveniens. This section governs Disputes between you and us. It does not apply to, and does not limit, a right a data subject has against us under the Standard Contractual Clauses incorporated in the Data Processing Addendum, which are governed by their own jurisdiction terms.

Class-action waiver. YOU AND WE EACH AGREE THAT A DISPUTE MAY BE BROUGHT ONLY IN AN INDIVIDUAL CAPACITY, AND NOT AS A PLAINTIFF OR CLASS MEMBER IN A PURPORTED CLASS, COLLECTIVE, CONSOLIDATED OR REPRESENTATIVE PROCEEDING. The arbitrator may not consolidate more than one person’s or organization’s claims and may not otherwise preside over a class, collective, consolidated or representative proceeding, except for the Mass Filing procedure above, which is not such a proceeding.

Severability of this section. If a court holds the class-action waiver unenforceable as to a particular Dispute, then, for that Dispute only: its class, collective, consolidated or representative aspect is severed and may proceed in court, while any individual claim within it remains subject to arbitration under this section — unless a court additionally holds that the entire Dispute must proceed in court, in which case the requirement to arbitrate is unenforceable only as to that Dispute. A ruling affecting one Dispute, one claimant or one organization does not make the class-action waiver or the agreement to arbitrate unenforceable for any other Dispute, claimant or organization, and does not affect any other provision of these Terms.

Opting out. You may opt out of the agreement to arbitrate and the class-action waiver — and only those. The governing law, the courts identified above, and the small-claims and injunctive-relief provisions are unaffected and continue to apply either way. Opt out by emailing legal@inxo.io within 30 days of the date you first accept these Terms, from an address that identifies the accepting individual, stating your organization’s name, your account email, and that you are opting out of arbitration. If you opt out, Disputes are resolved in the courts identified above, and nothing else in these Terms changes. An opt-out binds the organization that accepted these Terms and covers every workspace and mailbox connection under its account.

12. General

Entire agreement. These Terms, the Data Processing Addendum, the AI disclosure, and any order form or authorship election referenced in them are the entire agreement between us about the Service, and supersede any earlier discussions or agreements about it. An additional or different term in your purchase order or similar document has no effect, even if we do not object to it.

Assignment. You may not assign or transfer these Terms, in whole or in part, without our prior written consent, except to a successor of all or substantially all of your business by merger, acquisition, or sale of assets, provided the successor is not a direct competitor of ours and assumes these Terms in writing. We may assign these Terms without your consent in connection with a merger, acquisition, reorganization, or sale of all or substantially all of our assets. These Terms bind and benefit permitted successors and assigns.

Notices. We may give you notice by email to the address on your account, or by posting in the Console; notice is effective when sent. A notice of a revision to these Terms, the AI disclosure, the Data Processing Addendum, the Privacy Policy, or the authorship election is given by email to every administrator on your account and by a banner posted in the Console, both on the same day, and the notice period §9 states (for the Data Processing Addendum, its own §13) runs from that day. You may give us notice at legal@inxo.io, including the notices §9 and §10 describe, except that the arbitration opt-out in §11 must be sent as that section describes.

No waiver; severability. A party’s failure to enforce a provision is not a waiver of it. If a provision of these Terms other than the class-action waiver in §11 is held unenforceable, the rest remains in force, and the unenforceable provision is replaced with one that most closely achieves its intent.

Survival. Sections 3 (authorship), 6 (intellectual property), 7 (acceptable use, as to conduct before termination), 10 (confidentiality, warranties, liability, force majeure, and the effect of termination), 11 (governing law and dispute resolution), and this Section 12 survive termination of these Terms, as do the Data Processing Addendum’s deletion and return provisions (its §7), along with any provision that by its nature should survive.

---

Sub-processors: see the sub-processor register. Data protection: see the Data Processing Addendum. How the Service uses AI: see the AI disclosure.