Security at Kira
Security isn't something we bolt on once a feature works, it's part of how Kira is designed and operated from the start. Every module, from the Email Agent reading your inbox to the Call Agent handling a live phone conversation, was built on the assumption that it would eventually be trusted with genuinely sensitive information, and designed accordingly, rather than retrofitted once that trust was already extended.
We encrypt your data, restrict and log every access, watch our systems around the clock, and hold our own practices against recognized industry standards. This page is meant to explain what those commitments actually mean in plain language, not just list them as bullet points you're expected to take on faith. If you want the full technical detail behind any of this, our Security & Compliance page goes deeper into certifications, sub-processors, and infrastructure specifics.
A Direct Answer on Encryption
We're often asked whether Kira uses end-to-end encryption, and we'd rather answer that directly than let it sit as an implied assumption people might get wrong in either direction.
It doesn't. Here's why, plainly. Features like searching across your notes, letting an AI agent draft a reply based on the actual content of your inbox, generating a call summary from a real conversation, or having support meaningfully help troubleshoot an issue, all require Kira to be able to process your data. True end-to-end encryption, where only you hold the keys, would make every one of those features technically impossible, because the system itself would have no way to read what it's supposed to be working with.
Instead, we encrypt your data in transit and at rest using strong, industry-standard methods. This means your data is genuinely protected from anyone intercepting it over the network, and protected from anyone accessing storage directly without proper authorization. But Kira itself retains the technical ability to process that data, which is exactly what allows the product to function the way it's designed to. Internal access to your data by our own team is limited strictly to what's actually needed for support or operations, every single access is logged, and role-based controls mean not everyone internally has the same reach into customer data.
Highlights
- Encryption in transit (TLS 1.2+) and at rest (AES-256).
- Least-privilege access, with SSO and MFA on eligible plans.
- Continuous monitoring, logging, and a tested incident-response plan.
- A privacy-first default: we don't train models on your content unless you opt in.
- Regular vulnerability scanning and periodic penetration testing.
How Your AI Agents Respect Your Permissions
An agent working on your behalf, whether that's the Email Agent drafting a reply, a Call Agent summarizing a conversation, or an integration pulling context from a connected tool, never sees or uses anything you don't already have access to yourself. Kira's agents operate strictly within your existing permissions, they don't grant themselves broader access simply because they're acting automatically rather than manually.
This matters specifically because automation can otherwise become a quiet way to expand access without anyone noticing. If an agent could reach further than the person who configured it, that would be a meaningful gap, not a convenience. We built this the other way around deliberately: automation should never be a backdoor to more access than a human user actually has.
What Happens When Something Goes Wrong
No system is perfect, and we'd rather describe honestly what happens when something does go wrong than pretend that possibility doesn't exist. We maintain a documented incident-response plan and a 24/7 on-call rotation specifically so that an issue surfacing outside normal business hours doesn't sit unaddressed until someone happens to notice it the next morning.
If a security incident is confirmed to affect your data, we notify affected customers without undue delay, and within whatever timeframe applicable law requires, we'd rather share incomplete information early than stay silent while we investigate fully, because the choice to notify shouldn't be about protecting our own reputation first, it should be about giving you the information you need to protect yourself. After any incident, we run a genuine internal review of what happened and why, and where a gap is found, that becomes a concrete change to how the system works, not just a lesson noted internally and forgotten.
How We Protect Your Data
| Layer | What We Do |
|---|---|
| Encryption | TLS 1.2+ in transit; AES-256 at rest; managed keys with rotation. |
| Access | Least-privilege, role-based access; SSO and MFA; every access logged. |
| Infrastructure | Runs on AWS, with production kept isolated from development. |
| Monitoring | Continuous logging, alerting, and 24/7 on-call response. |
| Data use | No model training on your content, unless you tell us otherwise. |
How We Vet the Providers We Rely On
Before we bring on any sub-processor, whether that's infrastructure, payment processing, or an AI model provider, we review their security and privacy practices directly, including relevant attestation reports, penetration test summaries, and completed security questionnaires appropriate to how critical that provider actually is to the service. We don't treat this as a one-time check performed once and filed away. Providers are reviewed on an ongoing basis, and significant public security events affecting a provider we rely on get assessed specifically for whether they create any risk to you, not just noted as industry news.
This matters because your data's security depends on more than just our own systems, it depends on the full chain of providers involved in actually delivering the service, and treating that chain as a single onboarding checkbox rather than an ongoing responsibility would be a real gap in practice, even if it looked fine on paper.
Certifications and Frameworks
Our current certification status, SOC 2, ISO/IEC 27001, GDPR, and CCPA/CPRA, is maintained on our Security & Compliance page, so you're always looking at the most current information in one place rather than a version that might have gone stale here. We'd rather point you to a single, actively maintained source than risk two pages quietly drifting out of sync with each other over time.
Sub-Processors and Data Residency
We rely on a small set of vetted providers, each bound by data-protection terms consistent with our obligations to you. Keeping this list genuinely short is a deliberate choice, every additional provider is another party with some level of access worth thinking carefully about, so we only add one when it's genuinely necessary to deliver the service. The current list, along with regional data-residency options, is maintained on the Security & Compliance page.
Responsible Disclosure
We welcome reports from security researchers, and we mean that as more than a formality. If you believe you've found a vulnerability, email support@kiraai.ai. We acknowledge every report promptly and won't pursue good-faith research conducted under our guidelines, we'd rather treat an independent researcher's find as a genuine gift to our security posture than a threat to manage defensively.
Standards and References
Frequently asked questions
Quick answers to common questions.
Eligible customers can request our security overview, sub-processor list, attestations, and DPA by emailing support@kiraai.ai.
No. We do not train models on your content unless you opt in, and never on business customer content.
No, and we'd rather be upfront about that than let you assume otherwise. We encrypt your data in transit and at rest, but Kira retains the ability to process your data, which is what makes features like search and AI agents actually work in the first place.
No. Agents operate strictly within your existing permissions and never see or access anything you couldn't already access yourself, automation never becomes a way to quietly expand access.
We notify affected customers without undue delay, within the timeframe applicable law requires, and we run a genuine internal review afterward. Where a real gap is found, it becomes a concrete fix, not just an internal lesson learned.
We prioritize the frameworks that actually matter to the regions and industries our customers operate in, rather than collecting badges for their own sake. Our current, honest status on each is maintained on the Security & Compliance page.
Email support@kiraai.ai with details. We acknowledge promptly and protect good-faith research conducted under our guidelines.
Depending on your plan and region, US or EU data residency may be available. See the Security & Compliance page for current options.