Service Level Agreement
What we commit to for paid plans, how we measure it, and what you get if we fall short. This SLA forms part of the Paid Service Terms.
Last updated: August 3, 2026 · Applies to paid plans from August 3, 2026.
1. Scope
We run the whole Service — free and paid alike — to the availability target in Section 2, and we measure it the same way for everyone.
Service credits are available on Business and Enterprise plans. On the Free and Pro plans there is no credit and no contractual availability commitment: we still publish the target, the measurement, and the incident practice below, and you can hold us to the description, but a shortfall does not entitle you to a credit. If that matters to you, it is one of the things a Business plan buys.
Why the line is there: a credit has to be measured, adjudicated, and paid, and on a low-priced monthly plan that process costs more than the credit is worth. We would rather commit where the commitment is meaningful than publish one we would resent honouring.
2. Availability Commitment
Monthly availability target: 99.5%
We target 99.5% availability for the Service in each calendar month — about 3 hours 40 minutes of unavailability allowed per month. If we fall below it, you can claim a service credit under Section 5.
“Unavailable” means what it sounds like: the Service does not respond, or responds with errors, such that you cannot sign in or start a captioning session. Degraded quality, slower-than-usual captions, or the failure of an individual feature while the Service as a whole is usable are not counted as unavailability.
3. How We Measure
- Availability is measured per calendar month as (total minutes − unavailable minutes) ÷ total minutes.
- Measurement is taken by an independent external monitor that checks the Service every 5 minutes from outside our infrastructure — not by our own servers.
- Unavailability begins at the first failed check and ends at the first successful one, so the smallest interval we can record is 5 minutes.
4. Exclusions
The following do not count as unavailability:
- Planned maintenance announced at least 24 hours in advance, up to 4 hours per month. We try to schedule it outside business hours in the regions where most sessions run.
- Failures of an upstream provider we depend on — speech processing or payments — where the failure is on their side. Failures of the infrastructure we run the Service on are ours and count as unavailability. We will tell you when this is the case rather than call it our own outage or hide behind it, and we publish who those providers are in the Privacy Policy.
- Anything outside our reasonable control: network failures between you and us, denial-of-service attacks, natural disasters, or acts of government.
- Your own environment — your device, browser, microphone, extensions, or local network — and use of a browser or platform we do not list as supported.
- Suspension of your account for a breach of the Terms of Service, and non-payment.
- Features we have labelled beta, preview, or experimental.
5. Service Credits
“That month’s fee” means the fee attributable to the affected month. For a yearly subscription, that is one twelfth of the yearly price. Credits apply to Business and Enterprise plans (Section 1).
A credit is deducted from your next payment. Because the seller of record issues your invoices (Section 2 of the Paid Service Terms), we arrange that deduction with it on your behalf. If your subscription ends before a credit can be applied, we refund the equivalent amount instead. Credits are not otherwise paid out in cash.
6. How to Claim
Email support@newey.ai within 30 days of the end of the month you are claiming for, saying which month it is. You do not need to supply logs or evidence — we hold the measurement and will check it. If the commitment was missed, we apply the credit; if it was not, we will show you the figure we measured.
7. Incidents and Communication
For an outage that lasts more than 30 minutes, we email subscribers on paid plans what is happening and update them until it is resolved. For a significant incident we publish a short account afterwards of what happened and what we changed. Security incidents follow their own notification path, described in the Privacy Policy.
8. Sole Remedy
Service credits are the only remedy this SLA provides for a missed availability commitment. That does not displace rights you have elsewhere: your statutory rights as a Consumer, your right to cancel, and the refund provisions in the Paid Service Terms all remain available, and nothing here limits liability that applicable law does not allow us to limit.
9. Changes to This SLA
We may amend this SLA. A change that lowers the commitment or reduces credits is announced at least 30 days in advance with individual notice, and does not apply to a period you have already paid for. Improvements and clarifications take effect on publication.
Version history: announced on August 1, 2026 and effective on publication. First published together with the Paid Service Terms; applies to paid plans from August 3, 2026.
August 2, 2026, effective on publication: service credits are limited to Business and Enterprise plans. As first published this morning they were offered on every paid plan, including Pro. We narrowed it the same day, before any customer subscription existed under the wider wording — the only subscription taken was our own test purchase. The availability target, the measurement, the exclusions, and the incident practice are unchanged and still apply to everyone.
August 2, 2026: paid plans opened on August 2 rather than August 3, so this SLA applies from that date. The line above is left as written — a date we announced is corrected by a new line, not by editing the old one.
August 3, 2026, effective on publication, and all three changes widen what we owe rather than narrow it. Failures of the infrastructure we run the Service on are no longer excluded as an upstream-provider failure — they are ours and count as unavailability. A service credit is deducted from your next payment, and if your subscription ends before it can be applied we refund the equivalent amount instead of letting it lapse. And for an outage over 30 minutes we email subscribers on paid plans and keep updating them, in place of a promise to post updates on a status page we do not operate.