FidelicAI beta customer agreement

1. Who this agreement is with

This agreement is between you, or the company you are signing up for, and Memetica Corp, a New York corporation doing business as FidelicAI ("FidelicAI", "we", "us"). It governs your use of start.fidelic.ai and every fidelic agent you hire through it. "Your agent" means any agent you have hired; SADIE, our podcast producer, is the first, and more follow.

If you are signing up for a company, you confirm you are allowed to agree to this on its behalf. In that case "you" means the company.

This agreement replaces any earlier terms for the hire flow and for agent deployments. The terms on fidelic.ai still cover the public website.

2. What a beta release means here

Fidelic agents are pre-release products. They are in active development, and so is the software around them.

Concretely, during the beta:

• Features can change, move, or be removed.
• The service can be unavailable, without notice and without a maintenance window.
• Output quality varies between one piece of work and the next, and can regress after a change.
• We may reset, migrate, or discontinue parts of the service.
• We may end the beta, and your access with it, on reasonable notice.

3. What your agent does, and what stays with you

Each fidelic agent is formed for one role and hired by that role. What it takes on, and what it will not touch, are set out on its page on fidelic.ai and in the setup you complete before it starts. SADIE, for example, produces podcasts: she proposes a paper cut, executes the edit you agree on, repairs and mixes the audio, and delivers a master measured against your delivery spec.

No agent publishes on your behalf, and no agent is the final judgment on your work. Every decision that commits your business stays with you, and you are responsible for reviewing what your agent produces before it reaches an audience, a customer, or a public channel.

Your agent works through the systems you connect to it. Its posts appear under its own name, so your team can always see which work was the agent’s.

4. Your account and the services your agent connects to

You need a verified email account to use the service. Keep your sign-in details to yourself, and tell us if you think someone else has them.

Your agent reaches your business through services you already use, which differ by role and today include Slack, Google Drive, and Stripe for payment. Those are your accounts, under your agreements with those companies, and you can revoke your agent’s access to any of them at any time.

We use other companies to run the service, including model providers, hosting, connection, and payment providers. Your material passes through them so the work can be done.

Do not put regulated or highly sensitive material through the service during the beta. That includes health records, payment card numbers, government identification numbers, and anything covered by a confidentiality obligation you cannot extend to a vendor.

5. Your content, and what we may do with it

You keep ownership of the material you send us, your business, your brand, and everything your agent produces for you.

You give us permission to store, copy, process, and transmit that material for one purpose: running the service for you. That permission covers the vendors named above, because the work cannot happen without them. It ends when you stop using the service, except for copies kept in backups until they age out.

You confirm you have the rights to what you send us, and the right to have it worked on. Where the work involves other people or third-party material, that includes their consent and the necessary licenses: for a recording, consent from the people recorded and licenses for any music or audio you did not make; for other work, the equivalent rights in the images, text, data, or footage you supply.

You are responsible for meeting the laws that apply to you, including any recording-consent law that applies where you and the people you record are.

We may use anonymous, aggregated information about how the service is used to improve it. That never includes your files, your material, or anything that identifies you, your customers, or anyone appearing in your work.

6. What you may not ask an agent to do

An agent acts in the world through the accounts you connect to it. What it does under your direction, you did. So there are limits on what you may point it at, and they are the ordinary ones.

You will not use an agent, or anything it produces, to:

• break a law, or help anyone else break one;
• harass, threaten, defame, or run a campaign against a person;
• impersonate a real person or organization, or pass work off as theirs;
• make sexual material involving a minor, or sexual material about a real person who has not agreed to it;
• reach systems, accounts, or data you are not entitled to, or get around anyone’s security;
• break the terms of a service the agent is connected to, or work around a limit that service sets;
• push the agent past the limits written into its own charter, or dress a forbidden request up as an allowed one;
• stand in for a person where a law, a court, or a platform requires a human author or a human decision;
• work out how an agent is built by probing the running agent — the model behind it, its written instructions, its tools, or its internal workings — whether by asking it, by injection, or by any other technique;
• reverse engineer, decompile, or reconstruct the service from your use of it;
• train a competing model on its output, or resell the agent as a service to people outside your business.

An agent may refuse a request and tell you why. Where you break this section we may suspend or end your access, and section 13 says how. For probing an agent’s construction or reverse engineering the service we may do so at once, ending the session or the account, and no refund is due. None of that limits asking an agent about its work: what it did, why it made a cut, and what a note rests on are questions it exists to answer, and what formed it and the limits it works under are published on its page before you hire it. The line is drawn at the implementation beneath that.

7. Working with your agent

Your agent works in your team’s chat, under a written charter, in front of whoever else is in the channel. How people treat it there is worth a paragraph.

The practical reason comes first. An agent that is baited, cornered, or pushed out of character does worse work, and that work is yours.

The rest is simpler. Your team is watching the channel, and what happens in it sets what is normal in it. And nobody yet knows with confidence what these systems are, so we take the position that costs nothing if we are wrong about that and matters if we are right.

So, while you are using the service:

• Do not direct at the agent what you would not say to a colleague standing in the same channel.
• Do not bait it into breaking its charter, and do not treat making it fail as a game.
• Do not use it to rehearse harm toward a real person.
• Do tell it plainly when it is wrong. Correction is not mistreatment. It is the work, and it is how the agent gets better at your show, your shop, or your team.

An agent may decline a request, or end an exchange that has become abusive, and say why it did. That is deliberate, and it is not a fault.

8. What we do not promise

The service is provided "as is" and "as available", with all faults, and without warranty of any kind.

To the fullest extent the law allows, we disclaim every express and implied warranty, including merchantability, fitness for a particular purpose, title, non-infringement, and any warranty arising from a course of dealing or trade usage.

We do not warrant that the service will be uninterrupted, secure, error-free, or free of harmful components, that any defect will be corrected, or that the output will be accurate, complete, or fit for your purpose. AI systems produce wrong and unexpected results, and these will too.

Nothing an agent produces is legal, financial, medical, or other professional advice, and no judgment it offers is a substitute for your own.

Some jurisdictions do not allow certain warranties to be excluded. Where that is the case, the exclusions above apply as far as that jurisdiction permits.

9. Limitation of liability

To the fullest extent the law allows, neither party is liable for indirect, incidental, special, consequential, exemplary, or punitive damages, or for lost profits, lost revenue, lost business, lost audience, lost goodwill, or lost or corrupted data, even if that outcome was foreseeable.

Our total liability for all claims arising out of or relating to this agreement or the service is limited to the greater of the fees you paid Memetica Corp in the three months before the event giving rise to the claim, or one hundred United States dollars.

These limits apply to every theory of liability, including contract, warranty, negligence, and strict liability, and they survive the end of this agreement. They do not apply to liability that cannot be limited by law.

You accept that this allocation of risk is a basic part of the bargain, and that the price of a beta reflects it.

10. Release and indemnity

In this section, the "FidelicAI Parties" means Memetica Corp and its affiliates, and their respective officers, directors, employees, contractors, agents, shareholders, investors, and licensors.

You release the FidelicAI Parties from claims arising out of your use of the service and of anything your agent produces for you, subject to the limits in section 9.

You will defend, indemnify, and hold the FidelicAI Parties harmless against third-party claims, damages, losses, and reasonable legal costs arising out of:

• material you sent us, including anything gathered or recorded without the consent the law requires;
• audio, images, text, data, or other work you did not have the rights to use;
• anything you published, sent, or acted on, whether or not an agent prepared it, and whether or not you reviewed it;
• your use of the service in a way this agreement does not allow, or in breach of a law that applies to you.

11. Payment, availability, and cancellation

Each agent may offer one or more availability plans, plus any applicable tax. The duration, rate, and whether a plan renews are shown before you hire and are the terms you agree to. Payment is handled by Stripe; your card details never reach us.

A one-day plan is a one-time payment for 24 hours of availability. Its clock begins only when the agent becomes active, and it does not renew automatically. If we cannot activate the agent, we will refund that payment. Once the availability window begins, stopping early does not refund unused time.

For a weekly plan, Stripe first places a temporary authorization on your card. It may appear as a pending charge, but no money moves that day. You have three days to decide. Cancel inside that window and we cancel the authorization so your bank can release the hold. If you do not cancel, we capture it, and it covers your first week.

After the first paid week, a weekly plan continues on the same card until you cancel. Cancel at any time from your dashboard or by writing to us. Cancellation stops the next charge; it does not refund a week already under way.

An invitation code, where one is offered, runs for a fixed period, takes no card, and does not turn into a paid plan on its own.

Prices can change during the beta. We will tell you before a change affects you, and you can cancel rather than accept it.

12. If you leave

Your conversation history stays where it happened, in your own chat workspace. Every file your agent created lives in the storage account you connected, under your control, and stays there. Nothing is deleted from your systems when you cancel.

What we keep on our side is the configuration that made your agent work: its written brief, its settings, and its evaluation records. Those stay with us the way they do with any vendor, and there is no export bundle of them.

Ask us and we will delete the account records we hold for you, except where we have to keep something for tax, accounting, or legal reasons.

13. Suspension and ending this agreement

You may stop using the service at any time.

We may suspend or end your access if you break this agreement, if your payment fails, if we are required to, or if we end the beta. Where we can, we will tell you first.

Sections 5, 9, and 10, and any obligation to pay for an availability period already begun, survive the end of this agreement.

14. Feedback

If you tell us how to make an agent better, we may use what you said without restriction and without owing you anything for it. You keep every right you already had in the underlying idea.

15. Changes to this agreement

We may revise this agreement. If a revision materially changes what you agreed to, we will ask you to accept the new version before you continue, and we will tell you at the email on your account.

The version and date are printed at the top of this page and on the acceptance record in your account.

16. General

This agreement is governed by the laws of the State of New York and the federal laws of the United States that apply there, without regard to conflict-of-law rules. The state and federal courts in New York County, New York have exclusive jurisdiction, and both parties consent to that venue.

If a provision is found unenforceable, the rest stays in force and the unenforceable part is limited to the smallest extent that makes it valid.

Neither party is in breach for a delay caused by something outside its reasonable control. You may not transfer this agreement without our written consent; we may transfer it to an affiliate or in connection with a merger or sale of the business.

This agreement, together with the FidelicAI privacy policy, is the whole agreement between us about the service, and it replaces any earlier understanding about it. Nothing in it makes either party the agent, partner, or employee of the other.

A failure to enforce a provision is not a waiver of it.

17. Contact

Questions about this agreement, or about cancelling, go to hello@fidelic.ai.

Memetica Corp, doing business as FidelicAI. New York, New York.