Security

1. Every workspace is walled off

Everything a workspace owns — agents, websites, sources, indexed content, conversations and settings — is scoped to that workspace, and the database enforces this with row-level security. The public chat API works out which workspace a request belongs to from a verified key, never from anything the browser sends.

2. Where your data is

Our database, application servers and indexing worker run in Sydney, Australia. Some providers, including our AI model provider, process data in other countries; our Subprocessors page lists them. Data is encrypted in transit, and at rest by our database provider.

3. Keys and credentials

• API keys are stored as SHA-256 hashes, and secret keys are shown only once.
• Public widget keys work only on the domains you approve.
• You can disable or rotate any key at any time.
• Credentials you give us for integrations, calendars, actions and identity signing are encrypted with AES-256-GCM, using a key kept outside the database, and are never sent to a browser.

4. Accounts and staff access

• Every account must confirm its email address. Passwords must be at least 8 characters.
• Workspace roles control who can change settings, billing and team members.
• Our staff roles follow least privilege. Administrative actions are recorded in an audit log, and multi-factor authentication is available to staff.

5. Abuse limits

• Rate limits per key, IP address and visitor, plus limits per conversation.
• A maximum message length.
• Monthly usage allowances on every plan, with no charges for usage above them.

6. Calling other systems safely

• The crawler and actions refuse private network and cloud metadata addresses, so they cannot be pointed at internal systems.
• Webhook and API addresses and credentials never reach the visitor's browser, and a credential is never sent to a different host.
• An action that changes data runs at most once, and only after any confirmation you require.

7. Prompt-injection defences

Text retrieved from your content is treated as untrusted data. The assistant is instructed never to follow instructions found inside indexed content, and never to reveal its system prompt. These defences reduce the risk of manipulation; no AI system can rule it out entirely, which is why actions that change data can require confirmation.

8. Your data stays yours

We do not use your content or your customers' conversations to train AI models. Deleting a source removes it from the search index; deleting an agent also deletes its conversations. Voice audio is never sent to us.

9. What we don't claim

We do not currently hold SOC 2, ISO 27001 or other security certifications, and we do not offer an uptime guarantee. No system is completely secure. We will tell you plainly if that changes.

10. Reporting a vulnerability

If you find a security issue, email support@webanswerly.com with "Security" in the subject. Please give us a reasonable chance to fix it before disclosing it, and do not access other customers' data or disrupt the service while testing. We'll acknowledge your report and keep you updated while we fix it.