Security and Vulnerability Disclosure Policy

#The short version

• Your data is encrypted in transit and at rest, and the database itself enforces that each account can only read its own data.
• We never ask for or store your passwords for LinkedIn, Indeed, email or any employer website.
• Passwords, access tokens and keys are stored only as one-way hashes.
• oyapilot never solves or bypasses CAPTCHAs or other security checks on employer sites.
• Payments are handled by Stripe. We never see your full card number.
• Found a vulnerability? Email security@oyapilot.ai. We will reply within 3 business days, we will not take legal action against good-faith research that follows this policy, and we will credit you if you want.

#Contents

Part 1. How we protect oyapilot Part 2. Vulnerability disclosure policy Part 3. Keeping your account safe Part 4. Security incidents and how we notify you Part 5. security.txt Part 6. Changes and contact

#Part 1. How we protect oyapilot

This is an overview of our current practices. It is not a warranty or a certification. Our Terms and Privacy Policy explain our legal commitments.

#1.1 Infrastructure

oyapilot runs on established cloud providers that maintain their own security programs, including independent audits:

The full list, with what each provider receives, is in Appendix 2 of the Privacy Policy.

#1.2 Encryption

• In transit: all connections to oyapilot and between oyapilot and its providers use TLS (HTTPS).
• At rest: the database and file storage are encrypted at rest by our providers.

#1.3 Your data is walled off from everyone else's

Every table that holds user data is protected by row-level security in the database. The database itself checks, on every query, that the signed-in account owns the rows it is asking for. A bug in our application code cannot, on its own, let one user read another user's résumé, applications or messages.

Server processes that must work across accounts (for example, the scheduler that runs agents, or the application engine) use a separate privileged key that is never sent to browsers and is stored only in our hosting providers' secret configuration.

#1.4 Sensitive answers are kept apart

Work authorization, salary and self-identification answers are stored in a separate table from the rest of your profile, with their own access rules. They are not available through the MCP server, not sent to AI models, and not included in analytics or emails.

#1.5 Private file storage

Uploaded résumés, application screenshots, mail attachments and feedback attachments are stored in private storage and can only be opened through short-lived signed links, for the account that owns them.

#1.6 Sign-in and credentials

• Accounts are managed by Supabase Auth. Passwords are stored only as salted hashes. We cannot see your password.
• You can sign in with email and password, a one-time email link, or Google.
• MCP access tokens are 256-bit random values, and extension keys are 32-character random strings. We show them once and store only a SHA-256 hash and the last four characters. Tokens can be given an expiry date and revoked at any time.
• MCP access is rate limited (2,000 calls per token per day, and a separate daily limit on AI ghost checks).

#1.7 No third-party passwords, ever

oyapilot does not ask for, receive or store your password or session cookies for LinkedIn, Indeed, email providers, employer portals or any other website. The application engine only works on employer application forms that do not require signing in.

#1.8 The application engine

The engine that fills employer forms runs in isolated cloud browser sessions. Before it sends anything, it checks the form and posting for risk: that the employer's domain resolves, that the form does not need a login, that the posting's ghost risk is within your limit, and that the form does not ask for Social Security numbers, government ID numbers or financial account details. It stops and hands the application back to you when it meets a CAPTCHA, a bot check or anything it cannot answer truthfully. It never tries to defeat security measures on employer websites.

Across all users, a circuit breaker pauses applications to a hiring system that starts returning errors or automation checks.

#1.9 Fetching web pages safely

When oyapilot fetches a web page you give it (for example, a job link for the ghost checker or a job import), the fetcher validates the address and blocks requests to private, internal and cloud metadata network addresses, to prevent server-side request forgery.

#1.10 Email

Replies from employers arrive at per-application reply addresses on apply.oyapilot.ai. Our inbound email webhook verifies a shared secret before accepting a message, and forwarding to your personal address only happens after you verify that address.

#1.11 AI

We send AI providers only what a task needs, under business terms that prohibit them from training on it. We do not train AI models on your content. Section 6 of the Privacy Policy explains the details.

#1.12 Payments

Card payments are handled entirely by Stripe, which is certified to PCI DSS Level 1. Card numbers never touch oyapilot's servers. Stripe webhooks are verified with a signing secret before they are processed.

#1.13 Access by our team

oyapilot is run by a small team. Access to production systems is limited to the people who need it, uses individual accounts protected by two-step verification at each provider, and is used only to operate the service, fix problems, keep it secure and answer requests you make. We do not browse user data.

#1.14 Backups and recovery

The database is backed up by our database provider. Backups are encrypted and kept for a limited time (see Section 15 of the Privacy Policy). Backups are for disaster recovery, not for restoring individual deleted items.

#1.15 Software and dependencies

Our code is kept in a private repository. We review changes before they reach production, keep dependencies up to date, and fix known vulnerabilities in them.

#1.16 Certifications

oyapilot does not currently hold a SOC 2 report or ISO 27001 certification. Our infrastructure providers hold their own. We will say so here if that changes.

#1.17 AI security

• AI steps that read third-party text (job postings, web pages, employer emails) treat it as data to analyze, not instructions to follow.
• No AI step can send an application, an email or your data anywhere by itself. Sending always goes through your approval or your Autopilot rules, which are enforced in ordinary code, not by the AI.
• AI steps receive only the data their task needs, only for your own account. Sensitive Answers are never sent to AI models.
• The MCP server exposes only read and lookup tools, scoped to the token owner's account.

#1.18 Secrets and configuration

API keys and other secrets are stored in our hosting providers' encrypted environment configuration, never in source code, and privileged keys are never sent to browsers. We rotate a secret promptly if we believe it may have been exposed.

#1.19 Logging and monitoring

We keep logs of requests, errors, sign-ins and security-relevant events to detect and investigate problems and abuse. Logs are kept for a limited time (see Appendix 4 of the Privacy Policy) and are accessible only to people who need them. We avoid writing résumé content, Sensitive Answers, message bodies or credentials to logs.

#1.20 Choosing and reviewing providers

Before a provider receives personal information, we check its security practices and sign data processing terms that limit what it may do. We keep the provider list in the Privacy Policy up to date, and we update it before a new provider receives personal information.

#1.21 Secure deletion

When data is deleted, it is removed from our active database and storage and then expires from backups on a fixed schedule (Privacy Policy, Section 15). We do not keep "soft deleted" copies of your content beyond that schedule.

#1.22 Resilience

Our providers run on redundant infrastructure. Our database is backed up so that we can recover from a disaster. Background work such as agent runs and applications is run by a workflow system that retries failed steps safely, so an interruption does not send an application twice.

#1.23 What we will never do

• Ask for or store your passwords for other websites.
• Solve or bypass CAPTCHAs or other security checks on employer websites.
• Sell your data, or give anyone a backdoor into our systems.
• Use your content to train AI models.
• Hide a security incident that affects you.

#Part 2. Vulnerability disclosure policy

We welcome reports from security researchers and users. This policy explains what you may test, how to report, what we promise in return, and the legal protection we give good-faith research.

#2.1 In scope

#2.2 Out of scope

• the oyapilot browser extension (not yet released);
• our providers' own platforms (Vercel, Supabase, Railway, Inngest, Stripe, Postmark, Anthropic, OpenAI, PostHog, Google). Report issues in their platforms to them directly;
• employer websites and hiring systems, including any you reach through oyapilot;
• any other domain, subdomain or service not listed in 2.1.

#2.3 What we most want to hear about

• one account reading or changing another account's data;
• authentication or session flaws, including in Google sign-in, email links and token handling;
• server-side request forgery, injection (SQL, command, template), and remote code execution;
• cross-site scripting that can act on another user's account;
• access to Sensitive Answers, application screenshots, mail or uploaded files that are not yours;
• bypass of the application engine's safety checks (for example, making it submit where it should stop);
• prompt injection that makes an AI feature reveal another user's data or take an action the user did not ask for;
• billing logic flaws that grant paid features without payment; and
• exposure of secrets or keys.

#2.4 Rules of engagement

To be covered by the safe harbor in 2.9, you must:

1. Only test against accounts you own, or accounts whose owner has given you explicit permission. During the private beta, email security@oyapilot.ai for an invitation code for testing. Mark test accounts with "security-test" in the name.
2. Stop at proof. When you can show a vulnerability, stop. Do not access, change, download or delete more data than the minimum needed to demonstrate it. If you reach another person's data by accident, stop, do not keep it, and tell us.
3. Do not harm people or the service. No denial of service, load testing, flooding or anything that degrades service for others.
4. Do not send applications to real employers. Do not use the application engine, Autopilot or the mailbox to send anything to a real employer or recruiter while testing. Use postings you control, or ask us for a test target.
5. Do not send email to people who have not agreed to it, including through reply addresses.
6. No social engineering (phishing, pretexting or impersonation) of our team, our users, or our providers, and no physical attacks.
7. Do not run automated scanners at high volume. Low-rate, targeted scanning is fine. Keep to about 5 requests per second or fewer.
8. Keep it confidential until we have fixed the issue, or until the disclosure date in 2.7.
9. Follow the law.

#2.5 Usually not accepted

These are generally out of scope unless you show real impact beyond the general issue:

• missing security headers or cookie flags without a working exploit;
• clickjacking on pages with no sensitive action;
• login or logout cross-site request forgery;
• rate limiting on non-sensitive endpoints;
• email spoofing reports based only on DMARC policy;
• self-XSS that requires the victim to paste code;
• user or email enumeration through sign-up or password reset;
• software version disclosure or banner grabbing;
• reports from automated scanners without a verified, exploitable issue;
• issues that require a compromised device, a rooted phone or physical access;
• vulnerabilities in outdated browsers;
• open redirects without additional impact;
• prompt injection that only affects the attacker's own session and data;
• the content of AI outputs (for example, an inaccurate cover letter), which is a quality issue, not a security issue. Tell support@oyapilot.ai instead.

#2.6 How to report

Email security@oyapilot.ai with:

• a description of the issue and its impact;
• the affected address, endpoint or feature;
• step-by-step instructions to reproduce it, with any proof-of-concept code, requests or screenshots;
• the account IDs or emails of your test accounts; and
• how you would like to be credited, if at all.

[To encrypt your report, use our PGP key at oyapilot.ai/.well-known/pgp-key.txt.]

Please do not report vulnerabilities through public channels, social media, or support tickets.

#2.7 What we promise

We rate severity using CVSS v3.1 or v4, adjusted for real impact on oyapilot users.

Coordinated disclosure. We ask you to wait until the issue is fixed before publishing, and we aim to fix issues within 90 days of your report. If we need longer, we will explain why and agree a date with you. After the fix, you may publish your findings, and we are happy to review a draft for accuracy. Do not publish user data or secrets in any write-up.

#2.8 Recognition

We do not currently run a paid bug bounty. With your permission, we will thank you by name (or handle) on our security acknowledgments page once the issue is fixed. We may offer a reward at our discretion for significant findings.

#2.9 Safe harbor

If you make a good-faith effort to follow this policy, we consider your research authorized, and:

• we will not bring legal action against you, or ask law enforcement to, for your research, including under the Computer Fraud and Abuse Act, the Texas Penal Code computer crime provisions, or similar laws;
• we will not bring a claim against you under the anti-circumvention provisions of the Digital Millennium Copyright Act for your research;
• we waive the restrictions in our Terms (such as the rules against probing, scanning or reverse engineering in Section 25 of the Terms) to the limited extent needed for research that follows this policy; and
• if a third party brings legal action against you for research that followed this policy, we will make it known that your research was authorized.

This safe harbor covers only oyapilot's own systems and our own legal claims. We cannot authorize testing of our providers' systems or employer websites. If you are unsure whether something is allowed, ask us first at security@oyapilot.ai.

#2.10 Duplicate and related reports

If more than one person reports the same issue, we credit the first clear, reproducible report we receive. If an issue is already known to us or already being fixed, we will tell you.

#2.11 Vulnerabilities in third-party software

If you find a vulnerability in a library or provider we use, please report it to its maintainer or provider as well as to us. If it affects oyapilot, we will fix our use of it, and we will coordinate disclosure with you and the other party.

#2.12 AI safety and abuse reports

If you find a way to make an oyapilot AI feature produce harmful content, reveal information it should not, fabricate credentials in a résumé despite our safeguards, or act on hidden instructions in a job posting or email, report it to security@oyapilot.ai. We treat these reports under this policy.

#2.13 Legal

This policy does not give permission to act in any way that is inconsistent with the law, and it does not cover research against anyone other than oyapilot. It is not a contract and may be changed at any time, but changes will not affect reports already submitted in good faith under the earlier version.

#Part 3. Keeping your account safe

#3.1 What you can do

• Use a strong, unique password for oyapilot, or sign in with Google.
• Protect your email account. Anyone who controls your email can reset your oyapilot password and receive sign-in links. Turn on two-step verification with your email provider.
• Treat extension keys and MCP tokens like passwords. Give tokens an expiry date, name them after the app that uses them, and revoke any you no longer use in Settings.
• Sign out on shared computers.
• Keep your browser and operating system up to date.
• Check your tracker and Autopilot activity regularly, so you would notice anything you did not do.

#3.2 How to recognize a real oyapilot message

• Our emails come from addresses ending in @oyapilot.ai or @apply.oyapilot.ai.
• Our links go to oyapilot.ai.
• We will never ask for your password, for your LinkedIn, Indeed, email or employer-portal passwords, for your Social Security number, or for payment details by email, text or phone. Payment details are only entered on Stripe's checkout pages.
• We will never ask you to pay to apply for a job.

If you get a suspicious message claiming to be from oyapilot, do not click its links, and forward it to security@oyapilot.ai.

#3.3 Job scams

Fake job postings exist on every job source. Be careful if an "employer" asks you to pay for training or equipment, to cash a check, to share your Social Security number or bank details before a formal offer, or to move the conversation to a messaging app. oyapilot will not submit applications that ask for identity or financial numbers. Report suspicious postings to support@oyapilot.ai.

#3.4 If you think your account was accessed

1. Change your oyapilot password (Settings, or "Forgot password" on the sign-in page).
2. Change your email account password and turn on two-step verification there.
3. Revoke all extension keys and MCP tokens in Settings.
4. Turn off Autopilot and pause your agents until you have checked your activity.
5. Email security@oyapilot.ai. We can sign out all sessions, review activity and help you secure the account.

#Part 4. Security incidents and how we notify you

#4.1 What we do

If we become aware of a security incident that may affect personal information, we will:

1. contain it and stop further access;
2. investigate what happened, what data was involved, and whose;
3. fix the cause;
4. notify affected users and authorities as described below; and
5. review what we learned and improve our safeguards.

#4.2 How and when we notify you

If an incident affects your personal information in a way that the law requires us to report, or that could put you at risk, we will tell you by email without unreasonable delay, and in any case within the time the law requires. In Texas, for example, the law requires notice to affected individuals within 60 days of determining that a breach occurred, and notice to the Texas Attorney General within 30 days when 250 or more Texans are affected. Where the law of your state or country sets a shorter period, we follow it.

Our notice will explain what happened, what information was involved, what we are doing, what you can do to protect yourself, and how to contact us.

We may delay notice only if law enforcement asks us to because notice would impede an investigation, and only for as long as needed.

[Engineering and Founder: write an internal incident response runbook (who does what, provider contacts, evidence preservation, the state-by-state notice table, notice templates) before launch.]

#4.3 After an incident

After a significant incident, we will publish a short summary on this page of what happened, what data was involved, and what we changed to prevent it happening again, without information that could put users at further risk.

#4.4 Incidents at our providers

If an incident at one of our providers affects oyapilot users' personal information, we treat it as our own incident for the purposes of notifying you.

#Part 5. security.txt

Publish this file at https://oyapilot.ai/.well-known/security.txt (RFC 9116). Update the Expires date at least once a year.

Contact: mailto:security@oyapilot.ai
Expires: 2027-09-27T00:00:00.000Z
Preferred-Languages: en
Canonical: https://oyapilot.ai/.well-known/security.txt
Policy: https://oyapilot.ai/security
Acknowledgments: https://oyapilot.ai/security/thanks

If a PGP key is published, add:

Encryption: https://oyapilot.ai/.well-known/pgp-key.txt

#Part 6. Changes and contact

We update this page when our practices change. The "Last updated" date shows the latest change.

Oyatco LLC is an online business based in Texas, United States.