Questions and answers
Straight answers to what security, IT, and compliance teams ask about Zotniq.
About Zotniq
What Zotniq is, who it is for, and how it fits the AI-security landscape.
What is Zotniq and what problem does it solve?
Zotniq is runtime AI data protection. It inspects every prompt on the device before it reaches ChatGPT, Claude, Copilot, Cursor, or any other AI tool. Sensitive data such as PII, PHI, credentials, and source code is blocked, masked, or allowed based on your team's rules, in single-digit milliseconds.
What exactly counts as "AI usage" that Zotniq can see?
Zotniq covers six surfaces: browser AI (ChatGPT, Claude, Gemini, Copilot tabs), native AI apps (Claude Desktop, ChatGPT Desktop, Slack AI, Notion AI), coding assistants (Cursor, Copilot, Claude Code), application AI (features you ship), autonomous agents, and MCP tool calls. Personal browsing and non-AI apps are not inspected.
Who is Zotniq built for?
Security, IT, and compliance owners at teams from three users up to enterprise scale. Typical triggers: a SOC 2 audit asking about AI usage, HIPAA or GDPR obligations, blocking secrets from coding assistants, or discovering which AI tools employees already use before writing an acceptable-use policy.
How is Zotniq different from our existing DLP, EDR, or CASB tools?
Firewall sees the connection. DLP sees the file. CASB sees the app. Only Zotniq sees the prompt. Traditional tools go blind on TLS-pinned native AI apps, home wifi, and streaming responses. Zotniq inspects at the OS layer on the endpoint, before data leaves, and runs alongside your existing stack.
Is there a free trial or a free tier?
Yes. Zotniq offers a 14-day free trial with no credit card required. You can start in monitor-only mode to discover which AI tools your team uses and what leaves, then flip on enforcement when you're ready. Deployment on one laptop takes about five minutes to first finding.
How it works and deployment
How Zotniq installs, integrates with your stack, and what happens when policies change.
How is Zotniq deployed and does it need to be installed on every device?
Coverage requires the agent on each user's laptop. Setup takes minutes on macOS: install the agent, approve the system extension, and Zotniq installs a per-device root certificate in the local keychain for on-device TLS inspection. Team-wide rollouts push via standard MDM (Jamf, Intune, Kandji).
Will Zotniq conflict with our existing EDR or MDM tools (CrowdStrike, Jamf, Intune)?
Zotniq runs alongside standard EDR (CrowdStrike, SentinelOne) and MDM (Jamf, Intune, Kandji) stacks without interference. On macOS, Zotniq uses a userland Network Extension rather than a kernel hook, so it doesn't compete with EDR. Compatibility notes are available on request.
Does the agent intercept encrypted (HTTPS/TLS) traffic? Does it install a root certificate?
Yes to both, on-device only. Zotniq installs a per-device root certificate in the local keychain during setup so a userland helper can inspect prompt content in plaintext. Plaintext never traverses your network. Standard AI apps like Claude Desktop, ChatGPT Desktop, Cursor, and Copilot work without any per-app configuration because they use the system root store.
How quickly do policy changes take effect once we update a rule?
Policy updates apply live without an agent restart. New or updated rules propagate to endpoint agents within seconds of publishing from the console. Enforcement decisions themselves run locally on-device with no round-trip to a cloud service. No user action is required.
What happens if the agent loses its connection: does it fail open or fail closed?
The agent fails closed by default: cached rules continue enforcing on-device, so protection stays on even without network. New rule changes queue and apply once connectivity returns. All decisions during the outage are recorded locally and stream to the audit log when the agent reconnects.
Which AI tools and models does Zotniq actually cover?
Zotniq covers Claude Desktop, ChatGPT Desktop, and browser AI (ChatGPT, Claude, Gemini, Copilot tabs) plus coding assistants (Cursor, Copilot). Coverage is by network flow through the endpoint's network extension, so any AI app using standard HTTPS is inspected against your rules. MCP servers appear in the AI inventory.
Does Zotniq work with AI agents and MCP tool calls, not just chat prompts?
MCP tool calls appear in the Zotniq inventory: which MCP servers your team uses, when they are invoked, and from which app. Backend agents and SDK-based flows are covered separately by the Python SDK, which wraps OpenAI and Anthropic clients and applies the same rule engine to the requests you send from your code.
Does Zotniq integrate with our SIEM or existing security stack?
Audit events export as JSON or CSV from the console for import into any SIEM. Field mappings for common detection frameworks (MITRE ATLAS, EU AI Act Article 12) are documented so events land with the right classifications. Custom fields can be added on request for internal correlation.
Can we use our own SSO or identity provider for access control?
Yes. Zotniq authenticates via Kinde with Google Workspace SSO supported today. Additional identity providers that speak SAML 2.0 or OIDC (Okta, Azure AD, and others) can be enabled on request. Access is role-based within each organization.
Data collection and privacy
Exactly what Zotniq collects, where it goes, how long it stays, and what happens if you cancel.
What information does Zotniq collect from my device?
Zotniq collects findings (data type detected, which AI app, timestamp) and enforcement decisions. Default storage mode sends findings metadata plus a redacted snippet of each finding to Zotniq cloud. Metadata-only mode strips the snippet, so no prompt content ever leaves the endpoint. Both modes are customer-selectable per organization.
Does Zotniq read everything I type, or only what I send to AI tools?
Only what you send to a covered AI destination. Zotniq inspects text at the moment it is about to leave for a browser AI tab, a native AI app, a coding assistant, or an MCP tool call. It does not log keystrokes, personal documents, or unrelated traffic.
Is my data processed locally on my laptop, or sent to Zotniq's servers?
Detection runs locally on the endpoint. By default, findings metadata (data type, app, timestamp, decision) plus a redacted snippet of the finding leaves the device for the Zotniq cloud dashboard. Customers who want stricter data-sharing switch to metadata-only mode, where no prompt content or snippets ever leave the endpoint.
Where is our data stored and can we choose the region?
US and EU hosting regions are available. Region is set per organization at onboarding, and data does not cross regions once set. Full residency terms are documented in the DPA. Additional regions are added on customer request.
How long is our data retained, and can we control that?
Findings retention is customer-configurable per organization and set at onboarding based on your compliance framework. Backups rotate out on a documented schedule. Retention settings can be changed by an administrator at any time from the console. Full terms are in the DPA.
Does Zotniq use our prompts to train its AI models?
No. Zotniq's detection model is proprietary and pre-trained on synthetic and public data. It does not learn from customer prompts, findings, or metadata. This posture is documented on the Privacy page and enforced contractually in the DPA. Any future opt-in learning would require your express, revocable consent.
What happens to the data if we cancel our subscription?
On cancellation, Zotniq stops all data collection immediately. Customer data is deleted per the retention and deletion schedule documented in your DPA, and a deletion confirmation is provided in writing. Export tools let you take your findings and audit log with you as JSON or CSV before cancellation completes.
Will the agent slow down my laptop or browser?
No perceptible impact. Decisions run locally on-device with no round-trip to a cloud service. The agent uses a userland network extension rather than a kernel driver, so it doesn't compete with EDR kernel hooks or degrade system performance. Real-world benchmarks are available on request.
Rules and enforcement
How rules are scoped and what latency looks like when Zotniq is active.
Can we set different rules for different teams, roles, or data types?
Yes. Rules are scoped per team, per data type, and per destination. Defaults cover NAME, EMAIL, PHONE, credit card, SSN, IBAN, address, bank account, API secret, and PHI keywords, and customers add custom tags for anything else (source code, project names, internal identifiers). Engineering can allow one thing while finance blocks another, from the same console.
Will there be noticeable lag when I use an AI tool?
No. Decisions run locally on-device with no cloud round-trip. Users typing at normal speed report no perceptible difference between Zotniq-active and Zotniq-inactive sessions across ChatGPT, Claude, Cursor, and Copilot. Actual latency measurements per app are available on request.
Compliance
Which frameworks Zotniq maps to, current certification status, and how PII is handled.
What compliance frameworks does Zotniq support?
Zotniq maps to SOC 2, HIPAA, GDPR, UK GDPR, EU AI Act (Article 12 logging), ISO 27001, ISO 42001, PCI DSS, HITRUST CSF, GLBA (§501(b) Safeguards Rule), and SR 11-7. Full status per framework is on the Trust page. DPA and BAA are available at contract signing.
Is Zotniq SOC 2 certified today?
SOC 2 Type II is in progress with the report expected in 2027. Interim readiness documentation, control mapping, and a gap letter are available under NDA. Email [email protected] to request. HIPAA BAA and GDPR DPA are already available and signed at contract execution.
Will you sign a DPA or BAA with us?
Yes. The Zotniq Data Processing Agreement is published at zotniq.ai/legal/dpa and is accepted automatically as part of Terms acceptance at signup. Enterprise customers who need a wet-signed or mutually-negotiated DPA can request one via [email protected]. HIPAA Business Associate Agreements are signed via a separate sales-touch flow.
How does Zotniq handle PII or other sensitive personal data?
Two independent controls. Enforcement mode is per-rule: monitor-only records findings without blocking (discovery); enforcement blocks, masks, or allows before data reaches any AI destination. Storage mode is per-organization: default sends findings metadata plus a redacted snippet to Zotniq cloud; metadata-only mode ensures no prompt content ever leaves the endpoint.
Is Zotniq itself compliant with the EU AI Act as an AI system?
Zotniq's on-device detection uses classical machine learning and rule-based logic, not a general-purpose AI system. It falls outside the EU AI Act's high-risk obligations for us as the provider. For customers, Zotniq helps satisfy Article 12 (logging) and Article 26 (deployer obligations) for AI systems you use.
Question not answered here? Email [email protected] or book a demo.