Security & Trust
Business Friendly LLC · Last Updated: October 22, 2026
This page explains how Business Friendly LLC ("Business Friendly," "we," "our") protects the data our customers and their business communities trust us with on the BusinessFriendly.ai platform (the Business Friendly Index). It is a plain-language companion to our Privacy Policy and Terms of Service, not a replacement for either.
Access controls
Every organization on the platform is isolated from every other organization; server-side authorization checks run on every request, not just in the interface.
- Roles. Within an organization, an owner or admin manages who has access. Organization roles (owner, admin, staff) govern what a person can see and edit; confidential business records are further restricted to a named project team where the organization sets that.
- Passwords. Passwords are never stored in plain text. We store only a PBKDF2-SHA256 hash (100,000 iterations, a unique random salt per account) — we could not recover your password even if compelled to.
- Sign-in and sessions. Every sign-in is a session recorded on our side. You can see where you are signed in and end any session; organization owners and admins can sign a member out everywhere. Changing or resetting a password ends your other sessions. Failed sign-in attempts are rate-limited to slow down password-guessing.
- Multi-factor authentication (MFA). Organization accounts can enable MFA using an authenticator app plus one-time recovery codes, with a controlled process to reset MFA if a device is lost.
- Single sign-on (SSO). Organization accounts can connect their own identity provider over OpenID Connect — Microsoft Entra ID, Google Workspace, or another standards-based OIDC provider — and, where the organization chooses, require it for sign-in.
- Security Center. Organization admins get a Security Center covering active sessions, sign-in history, MFA and SSO status, connected integrations, organization API keys, and recent permission and sharing changes.
- Organization API keys. Keys are scoped to specific resources, named, revocable, and rate-limited. We store only a SHA-256 hash of the key — the full key is shown once, at creation, and cannot be retrieved again.
- Webhooks. Outbound webhook payloads are signed (HMAC-SHA256) so a receiving system can verify they came from us.
- Audit log. Consequential actions — sharing changes, exports, API key and webhook changes, permission and security-setting changes — are written to an append-only audit log that application code never edits or deletes. Sign-in attempts, successful or not, are kept in a separate sign-in history that organization admins can review.
Encryption
- In transit. All traffic to businessfriendly.ai runs over HTTPS (TLS) through Cloudflare.
- At rest. Application data lives in Cloudflare D1 (SQL) and Cloudflare R2 (file storage), both encrypted at rest by Cloudflare.
- Sensitive tokens. Where we hold a credential on a customer's behalf — for example, an OAuth token from a connected Microsoft 365 calendar, or a webhook signing secret — it is separately encrypted (AES-GCM) with a key held as a Cloudflare Workers secret, not in the database in the clear.
Data & AI trust
Our full Data & AI Trust Standard covers source provenance, confidence, human approval and confidentiality levels in detail. In short:
- Where the platform uses AI to draft an answer, summarize a record, or analyze a document, the relevant customer content is sent to Anthropic's Claude API to generate that response. We send only the content needed for that request.
- Business records marked business confidential are restricted to the organization (and, where set, a named project team) — that visibility level is enforced on the server for every read, including AI features, so confidential information stays where its owner put it.
- AI drafts and analyzes; it does not make final decisions. Adverse classifications of a business, pricing and product changes, and anything sent externally pass through a named human before they take effect.
- AI-generated text on the platform is labeled.
Backup & disaster recovery
Our database runs on Cloudflare D1, which provides Time Travel: point-in-time restore of the database to any point in the preceding 30 days. We rely on this platform capability for disaster recovery of application data. Uploaded files (documents, recordings, images) are stored in Cloudflare R2.
Incident response
If you believe you have found a security issue affecting Business Friendly, or you are a customer reporting a suspected incident, contact consulting@businessfriendly.ai. We investigate reported incidents, and where a confirmed security incident affects a customer's data, we notify the affected customer without undue delay with the details we have and what we are doing about it.
Vulnerability disclosure
We welcome good-faith reports of security vulnerabilities in Business Friendly.
- How to report. Email consulting@businessfriendly.ai with enough detail to reproduce the issue (affected URL, steps, and impact). Please do not include real customer or respondent data in your report.
- Safe harbor. If you make a good-faith effort to report a vulnerability to us privately, avoid accessing or modifying data beyond what is needed to demonstrate the issue, avoid degrading service for others, and give us a reasonable opportunity to fix the issue before any public disclosure, we will not pursue legal action against you for that research and will treat your report as authorized.
- What to expect. A security report is a Severity 1 request under our published
response commitments: we respond within one business
day — a human reply that acknowledges the report and says what happens next — and then work
continuously toward a workaround or a fix. We will tell you what we found and, where a change ships, that it
shipped. We do not currently operate a paid bug-bounty program.
Our machine-readable disclosure contact is published at /.well-known/security.txt per RFC 9116.
Data retention, export & deletion
- Retention. We retain account and survey/business data for as long as an account is active or as needed to provide the service. If you close your account, tell us and we delete or anonymize your personal information, normally within 30 days of your request, and confirm when it is done. Security and audit logs, billing records, and anything we must keep by law are retained longer — see our Privacy Policy for the full policy.
- Export. You are not locked in. Depending on your plan, you can export your own data as a report (PDF, Word, or CSV), export your business list as CSV, export a selected map cohort as CSV, or pull records programmatically through the organization API.
- Deletion requests. To request access to, correction of, or deletion of your personal information, contact consulting@businessfriendly.ai.
Subprocessors
We use a limited set of subprocessors — companies that process data on our behalf to provide the service. See the current list, with what each one does and what data it sees, on our Subprocessors page.
Accessibility
Our customers are local governments, and the reports this product generates end up in council
and board packets that those bodies may themselves be required to make accessible. We build to
WCAG 2.1 Level AA, and the report's charts carry text alternatives and data tables so a packet
built from our export carries the same information in a readable form. See our
Accessibility statement.
How this is enforced, not just intended
Most security pages describe policies. These are the mechanisms, and each one is enforced in
code and checked automatically before every release.
One community can never reach another's data
Every query is scoped to the organisation that owns the record, on the server — not in the
browser, where it could be bypassed. We verify this by attacking it: an automated test signs in
as one customer and attempts to reach another customer's survey, report, comments, exports,
segmentation and action plan, by reading and by writing. Every attempt must be refused, and the
test runs on every release. A build that leaks anything between customers does not ship.
Survey answers are separated from the people who gave them
This one is architectural rather than procedural. Ratings and written comments are stored
against a random identifier with no path back to a name or email address. It is
not that we choose not to look — the connection does not exist to follow. That is why we can
promise a business owner their individual answers stay confidential and mean it.
Small groups are protected from being identified
In a community of a few hundred businesses, "the three manufacturers rated permitting 1.9"
identifies three firms and their opinion of their own government. Results are withheld below a
minimum number of respondents — and when withholding one group would let a reader work it out by
subtracting the rest from the total, a second group is withheld with it. That protection applies
everywhere the numbers appear, including exports and anything our assistant can answer.
Written comments are reviewed before anyone — or anything — sees them
Email addresses, phone numbers, street addresses and links are removed from written comments
automatically, and anything that may name a person is flagged for a human. The same reviewed text
is what our AI features receive: the model is never given the raw comments. An instruction telling
a model to behave is not a control; not handing it the data is.
Our AI cannot invent a quote
Where the product shows a quotation from a comment, that text is checked word for word against
what was actually submitted before it can be displayed. Anything that does not match exactly is
discarded. Scores, charts and tables are calculated arithmetically and never depend on AI at all.
You control what becomes public, and can take it back
Nothing about your community is published until you publish it. When you withdraw that
consent, the public page stops existing for search engines, and your result is removed from the
peer comparisons other communities see — in the same action, not on a later cleanup.
Sign-in protection that fits your organisation
Two-factor authentication with an authenticator app and single-use recovery codes is available
to every account. Larger organisations can require single sign-on through their own identity
provider, enforced for their members, without imposing it on smaller customers. Sign-ins,
refusals and session revocations are logged, and you can sign out every other device at once.
A control program with dated evidence
We run a documented internal security program in which each control's status is derived from
dated evidence of it actually operating, rather than from a description of what is supposed to
happen; a control without that evidence is not counted as operating. Consequential changes are
recorded against the release that shipped them, and the evidence record itself cannot be edited
or deleted.
Have a specific requirement? Ask Ally, our
assistant, at the bottom right of any page — security questionnaires, procurement conditions,
data-handling rules your state or council requires, or how we would fit alongside a system you
already run. Ally answers directly from what this platform actually does. For anything else,
email consulting@businessfriendly.ai.