Ethics Policy

Effective date: 19 September 2026

Lara (also called LaraBot) at larabot.co is a personal expense companion in Telegram and WhatsApp, operated by Johnny Nassar, an individual (sole trader) trading as Lamma. This page explains the rules we bind ourselves to when we build, message, and experiment. It sits alongside our Privacy Policy and Terms of Service.

We never

• Use fake scarcity or urgency.
• Shame, guilt, or frighten you.
• Exploit financial stress.
• Build addictive streak or loss-aversion mechanics.
• Make misleading or unverifiable claims.
• Hide or complicate opt-out.
• Nudge spending that serves the business over you.

We always

• Speak plainly and honestly.
• Tell you at the start that Lara is an AI assistant, and which AI providers process your messages.
• Keep opt-out simple: reply stop (or unsubscribe) to turn off daily reminders, nudges minimal for fewer nudges, quote off for the daily quote, and delete my data to leave. recaps off stops the weekly summary and monthly recap (recaps on brings them back).
• Label financial information, not advice, with a clear disclaimer.
• Minimize and protect data (phone-hash + encryption) as described in the Privacy Policy.
• Comply with Telegram, WhatsApp/Twilio, advertising, and data-protection rules that apply to us.

How this is enforced today

We only list controls that exist in the product right now.

• No A/B experiments are running. Lara does not currently split users into test groups.
• Coaching messages adapt over time. Lara's weekly coaching learns which kinds of nudges people found useful. When the coaching rules are rewritten by one AI model, a second, independent AI check reviews the rewrite against this policy and rejects it if it breaks a rule; the previous rules then stay in force.
• Version history. Each accepted change to the coaching rules is saved as a new version rather than overwriting the old one.
• Monitoring. An automated monitor checks every few minutes for outages, error spikes, runaway costs, and unprotected phone numbers in the database, and alerts the founder. It alerts; it does not switch anything off by itself.

Planned, not yet built

• Asking for your explicit consent before including you in any A/B experiment. No experiment will run on users before this exists.
• Founder sign-off recorded for each experiment before it goes live.
• Automatic pause of a message type when user-harm signals appear (for example, a spike in opt-outs).
• A tamper-proof log of experiment decisions.

Coaching data & deletion

When Lara runs optional coaching or nudges, related behavioral rows are kept only as long as needed for that loop (on the order of ~12 weekly cycles), aligned with what the product is allowed to learn from. If you send delete my data and confirm, we anonymize coaching rows tied to you and delete raw AI outputs that may have embedded your spending details. Aggregate playbook stats may remain without your identity. Full account reset and remaining erasure follow the Privacy Policy (including email to johnnynassar@bylamma.com).

Contact

Questions about this policy: johnnynassar@bylamma.com. Product support: support@larabot.co. Email is our support channel. If Lara offers a support command in your chat, you can use that too.