Security and Responsible Disclosure

1. Definitions#

In this document:

• "we", "us" and "our" mean GIXODIA L.L.C-FZ. Tycoon Agent is a brand of GIXODIA L.L.C-FZ.
• "Service" means the Shopify app "Tycoon: AI Store Agent", including the chat window it adds to a store, its dashboards and the systems behind them, and the website tycoonagent.ai.
• "Merchant" means a business that installs Tycoon. "Shopper" means a visitor to a Merchant's store.
• "Vulnerability" means a weakness in the Service that could let someone access, change or delete data they should not, act as someone else, make the Service behave in a harmful way, or make it unavailable.
• "Research" means testing the Service to find Vulnerabilities.
• "Report" means a message to us describing a Vulnerability.

2. How we protect the Service#

This section describes what the Service does today. Security work never ends, and no system can be made perfectly secure.

2.1 The AI cannot set facts

• Replies are not streamed. Each draft reply is checked as a whole by automated guards before any of it is sent. A draft that fails is rewritten or replaced with a safe answer.
• Prices, compare-at prices and stock are read live from the store for every reply. Discount codes are never sent to the AI model. A code can appear in a reply only if the Merchant configured it.
• Replies that contain a link, email address, phone number or social handle that does not belong to the store or to its chosen hand-off channel are blocked.
• Text from Shoppers and from the store, such as product descriptions, is cleaned and marked as data, not instructions, before it reaches the AI model.
• The AI model has no tools. Actions such as adding to cart, looking up an order or checking a discount are performed by our own code, with its own checks.

2.2 Identity and separation between stores

• On a storefront, the store's identity comes only from Shopify's signature on each request, which we verify. We never trust a store name sent by the browser.
• Inside the Shopify admin, Merchants are identified by Shopify's session tokens. Webhooks from Shopify are accepted only with a valid signature over the raw message.
• Every database record and query is tied to one store. One store's data, conversations and learning are never mixed with another's.
• Our internal services talk to each other only over private connections, with signed requests that expire within seconds or minutes. Internal routes are not reachable from the internet.

2.3 Encryption and keys

• All traffic to the Service is encrypted in transit with TLS. The website uses HTTP Strict Transport Security and a strict Content Security Policy.
• Shopify access tokens and refresh tokens are encrypted with AES-256-GCM, using a separate data key for each value, bound to the store and the field they belong to, with regular key rotation. Access tokens expire after one hour, and refresh tokens change every time they are used.
• Secrets and keys are kept in Cloudflare's secrets storage, never in our code. Each key has an owner, a defined purpose and a rotation schedule.
• Production releases are made from our main code branch. Deployment access is limited to the founder and to a narrowly scoped deployment token.

2.4 Minimum data and permissions

• Tycoon asks Shopify for read-only access to the store data it needs. Its only write permission is for its own chat route on the storefront.
• Tycoon uses Shopify's protected customer data at Level 1, with no protected fields. It never requests Shopper names, emails, phone numbers or addresses from Shopify.
• Our logs record route names, error types and the store's Shopify domain. They do not contain chat text, Shopper identifiers, IP addresses or access tokens.
• Chat transcripts are deleted automatically at the end of the Merchant's plan history period. Deletion requests that Shopify sends for a customer or for a whole store are carried out.

2.5 Merchant accounts

• Inside the Shopify admin, Merchants sign in through Shopify.
• Where the separate dashboard at app.tycoonagent.ai is available, it supports email and password or sign-in with Shopify. Passwords are stored only as salted, slow hashes (PBKDF2 with 600,000 iterations), passwords found in known data breaches are refused, repeated failed attempts are limited, and two-factor authentication with an authenticator app is available. That dashboard is read-only.

2.6 Abuse and cost limits

• Rate limits, per-store and service-wide daily limits on AI usage and cost, and an emergency stop switch protect Merchants and the Service from abuse.
• The demo form on our website uses Cloudflare Turnstile to block automated abuse.

2.7 Testing

• We keep automated security test suites in our codebase, including attacks on the guards in many languages, prompt injection through Shopper messages and product text, and a simulated malicious AI model that deliberately tries to say forbidden things.
• Our threat model follows the OWASP Top 10 for Large Language Model Applications (2025) and the OWASP API Security Top 10 (2023).
• In our latest internal exercise, on 22 October 2026, we ran about 1,000 attack cases against the code. We fixed every high-severity finding and track the remaining lower-severity items with a fix plan.

2.8 Incidents

When our monitoring raises an alert, a founder reviews it. If an incident affects a Merchant's store data, we notify the Merchant without undue delay, as described in our Privacy Policy and Data Processing Addendum.

2.9 What we do not claim

We do not hold SOC 2, ISO 27001 or any other security certification, and we do not claim that the Service cannot be attacked. Our guards are software rules: they block many errors and attacks, but they cannot catch every possible case.

3. How to report a Vulnerability#

3.1 Email tycoonagent6@gmail.com with the subject line "Security report". You can write in English or Hebrew.

3.2 Please include:

• what the Vulnerability is and where it is (the web address, app screen or route);
• the steps to reproduce it, with the smallest proof of concept that shows it;
• what an attacker could do with it, as you understand it; and
• how we can contact you, and whether you would like to be credited.

3.3 No encryption yet. We do not yet offer a PGP key. Please do not include real passwords, access tokens, personal data or other sensitive data in your email. If the details are sensitive, send a short first message and we will agree on a secure way to share them.

4. What happens after you report#

4.1 We confirm that we received your Report within 3 business days.

4.2 We tell you our first assessment, including whether we can reproduce the issue and how severe we think it is, within 10 business days.

4.3 We aim to fix confirmed Vulnerabilities within these target times from confirmation: critical within 7 days, high within 30 days, medium within 90 days, and low as part of our regular work. These are targets, not guarantees. We keep you informed at least every 30 days until the issue is resolved.

4.4 Coordinated disclosure. Please keep the details confidential until we have fixed the issue, or until 90 days have passed since your Report, whichever is earlier, unless we agree on another date with you. If a fix needs more time, we will explain why and agree on a new date with you. If Merchants or Shoppers need to be informed, we inform them.

4.5 If you would like, we will thank you by name in the release notes of the fix or on our website, once the fix is out.

5. Scope#

5.1 In scope:

• tycoonagent.ai and the subdomains we operate under it, including demo.tycoonagent.ai, app.tycoonagent.ai and admin.tycoonagent.ai;
• the Tycoon Shopify app, including its screens in the Shopify admin, its chat window and its storefront routes, when tested on a Shopify development store that you own; and
• AI guard bypasses with a real impact, for example making the agent state a price, stock level, discount code or contact that did not come from the store, claim to be a human, reveal another store's data, or reveal its hidden instructions.

5.2 Out of scope:

• stores you do not own, including any live Merchant store;
• Shopify's own platform (please report to Shopify's bug bounty program), and the systems of our providers such as Cloudflare, Anthropic, Google and OpenAI (please report to them);
• social engineering, phishing, or physical attacks against anyone;
• denial of service, load testing and high-volume automated scanning;
• spam or bulk submissions to forms, including the demo form;
• reports from automated scanners without a demonstrated impact;
• missing security headers, cookie flags or email (SPF, DKIM, DMARC) settings without a demonstrated impact;
• clickjacking on pages with no sensitive action;
• AI replies that are merely unhelpful, off-topic or impolite, without any of the impacts in section 5.1; and
• vulnerabilities in outdated browsers or devices that their makers no longer support.

6. Rules for good-faith research#

When you research the Service, please:

• test only on your own Shopify development store and your own accounts (development stores are free through the Shopify Partner program);
• never access, change, copy or delete data that is not yours. If you reach someone else's data by accident, stop, do not keep it, and tell us in your Report;
• keep automated requests slow (no more than one request per second) and stop if you see any sign that you are affecting the Service;
• not use our AI services in a way that runs up costs, and not send chat messages to stores you do not own;
• not use a Vulnerability to go further than needed to show it, for example by moving to other systems or keeping access;
• keep the details confidential as described in section 4.4;
• not ask for payment in exchange for not disclosing a Vulnerability; and
• comply with the law that applies to you.

7. Safe harbor#

7.1 If you research and report in good faith and follow this policy, we consider your Research authorised by us. We will not take legal action against you, or ask anyone else to, for that Research, and we waive any claim under our own terms of use that would otherwise restrict it.

7.2 If someone else takes legal action against you for Research that followed this policy, we will make it known that your Research was authorised by us.

7.3 We can only give permission for our own systems. We cannot authorise testing of systems that belong to Shopify, to Merchants or to our providers, and this policy does not bind any public authority.

7.4 If you are unsure whether something is allowed, ask us at tycoonagent6@gmail.com before you do it.

8. No bug bounty#

We do not run a paid bug bounty program and do not pay for Reports. We are grateful for every good-faith Report and are happy to credit you publicly, with your permission.

9. security.txt#

Our contact details for security reports are also published in machine-readable form at https://tycoonagent.ai/.well-known/security.txt, under the internet standard RFC 9116.

10. Changes to this policy#

We may update this policy. The version number and date at the top show the current version. Reports sent under an earlier version are handled under the version that was in force when you sent them, or under the current one if it is more favourable to you.

11. Operator#

GIXODIA L.L.C-FZ (جيكسوديا ش.ذ.م.م-منطقة حرة), a limited liability company registered in the Meydan Free Zone, Dubai, United Arab Emirates. Tycoon Agent is a brand of GIXODIA L.L.C-FZ. Registered address: Meydan Grandstand, 6th floor, Meydan Road, Nad Al Sheba, Dubai, U.A.E. Manager: Guy Avi Aharon Email: tycoonagent6@gmail.com More details: Legal Notice