Hero background

Blue Machines Vulnerability Disclosure Policy

Introduction

Welcome to the Blue Machines AI Vulnerability Disclosure Program (VDP). Blue Machines AI is at the forefront of developing powerful, full-stack voice AI agent platforms. We are dedicated to pushing the boundaries of artificial intelligence, creating sophisticated models that understand and generate natural-sounding speech, enabling seamless human-AI interaction across various applications and industries. Our mission is to build accessible, efficient, and robust AI tools that empower businesses and developers to integrate human-like conversational capabilities into their products and services.

While our dedicated teams work tirelessly on the security, safety, and robustness of our core voice AI platform and associated services, we recognize that achieving truly airtight security requires diverse perspectives and expert scrutiny. That's why we created this Vulnerability Disclosure Program – a dedicated channel for security researchers and bug bounty hunters to responsibly contribute to the security of our AI infrastructure and platform.

We are constantly striving for excellence in the three pillars of our AI development: scalability, reliability, and security. We take all reported vulnerabilities extremely seriously. If you have the skills and dedication to help us identify potential weaknesses in our web platform, APIs, or infrastructure, we encourage you to read on and learn how you can become a crucial part of our security program.

In Scope

Web Applications: All publicly accessible web interfaces and applications provided by Blue Machines AI (bluemachines.ai and any subdomains).

APIs: All REST APIs and other programmable interfaces offered by Blue Machines AI.

Accepted Vulnerabilities

Blue Machines AI accepts responsible disclosure of any vulnerability directly impacting the systems in scope, provided it is unique and has more than a P5 impact according to Bugcrowd's VRT/Vulnerability Rating taxonomy. This includes AI-specific vulnerabilities relevant to large language models (LLMs) and AI systems, encompassing risks like prompt injection, data privacy issues, model stealing, adversarial attacks, and those listed in the OWASP Top 10 for LLM Applications, as well as traditional web/software vulnerabilities affecting our platform's infrastructure or services.

Out of Scope

  • Any domain not mentioned in the list of in-scope systems.
  • Third-party vendor/applications/services/platforms used by Blue Machines AI.
  • API key disclosure without proven business impact.
  • WordPress usernames disclosure.
  • Self-XSS that cannot be used to exploit other users.
  • Verbose messages/files/directory listings without disclosing sensitive information.
  • CORS misconfiguration on non-sensitive endpoints.
  • Missing cookie flags or security headers.
  • Cross-site Request Forgery (CSRF) with no or low impact.
  • Presence of autocomplete attributes on web forms.
  • Reverse tabnabbing.
  • Non-existence of rate limits.
  • Email bombing or best practices violations (password complexity, expiration, reuse, etc.).
  • Clickjacking without proven impact or unrealistic user interaction.
  • Sessions not being invalidated.
  • Tokens leaked to third parties.
  • Anything related to email spoofing (SPF, DMARC, DKIM).
  • Content injection without modifying HTML.
  • HTTP request smuggling without proven impact.
  • Homograph attacks.
  • Banner grabbing or version disclosure.
  • Not stripping metadata from images or files.
  • Same-site scripting (XSS) without impact.
  • Subdomain takeover without taking over the subdomain.
  • Origin IP disclosure.
  • Disclosed or misconfigured Google Maps API keys.
  • Host header injection without proven business impact.
  • Findings related to outdated Swagger versions or related vulnerabilities.
  • Cache poisoning or metrics exposure.
  • Vulnerabilities requiring extensive user interaction.
  • Vulnerabilities that require root access or bypass certificate pinning on rooted devices.
  • Attacks requiring physical access, social engineering, phishing, or other fraud activities.
  • AI-specific vulnerabilities with low impact, such as minor hallucinations or inconsistent outputs without proven harm.
  • Bypassing safety controls or ethical guidelines without causing unintended behavior or data leakage.
  • Data poisoning or training data manipulation without proven impact on model integrity or output.
  • Model stealing or extraction attempts without proven success or impact.
  • Adversarial attacks without demonstrated capability to manipulate model outputs or cause harm.
  • Prompt injection or jailbreaking attempts without bypassing safety mechanisms or causing unintended behavior.
  • Insecure deserialization or injection flaws targeting AI-specific input handling.
  • CWE-359 (exposure of sensitive configuration) related to AI model parameters or weights.
  • Cross-site Scripting (XSS) targeting AI output validation or sanitization.
  • Homograph attacks or typosquatting related to AI model inputs or outputs.
  • Banner grabbing or version disclosure of AI model libraries or frameworks.
  • Not stripping metadata from AI-generated content or files.
  • Same-site scripting (XSS) targeting AI output validation or sanitization.
  • Subdomain takeover without taking over the subdomain.
  • Origin IP disclosure.
  • Disclosed or misconfigured Google Maps API keys.
  • Host header injection without proven business impact.
  • Findings related to outdated Swagger versions or related vulnerabilities.
  • Cache poisoning or metrics exposure.
  • Vulnerabilities requiring extensive user interaction.
  • Vulnerabilities that require root access or bypass certificate pinning on rooted devices.
  • Attacks requiring physical access, social engineering, phishing, or other fraud activities.

Prohibited Testing Methodologies

  • Any type of Denial-of-Service (DoS) or Distributed Denial-of-Service (DDoS) attacks or testing with automated scanners.
  • Brute-forcing or dictionary type attacks against any system component.
  • Phishing or any type of social engineering techniques targeting users, employees, or researchers.
  • Attacks against the integrity of user data or employee data.
  • Attempts to compromise user accounts or employee accounts.
  • Modify, disrupt, or degrade the quality of Blue Machines' services or systems.
  • Conducting aggressive probing or reverse-engineering solely to identify potential jailbreak prompts or bypass methods without demonstrating actual harmful or unintended output.
  • Using adversarial attacks with the specific intent to cause catastrophic failure, misinterpretation leading to severe harm, or disable the AI's core functionality.

Reporting Procedure

The reporting procedure is fairly simplified for ease of use. In order to make a responsible disclosure conforming to the VDP, please fill the below form correctly and provide all the details for the fields marked by (*).

Report Form Link:

https://docs.google.com/forms/d/e/1FAIpQLScr11EwVlD6X-bitSc30ksb6SyC9R_z9tt8jGWfe3FFmeVjbw/viewform?usp=header

Be very careful while filling this form as you'll not be able to edit it later. In case you made any mistake then please submit another form with the "Is this VD a fixed version of a previously submitted disclosure which had errors?" field marked.

In scenarios where there is any new update regarding this vulnerability (whether it is a fix bypass or it is another exploitation method or a chained-vulnerability or a new impact related to this vulnerability) you must fill a new form.

If you do find any vulnerabilities that are associated with a third party Vendor/applications/services/platforms of Blue Machines, then immediately suspend all testing and inform us via Official Channel. We’ll get back to you in such cases, after taking the necessary steps to notify the 3rd party. If they allow further testing and our security team considers further testing to be necessary, we’ll inform you of the steps you need to follow then. That is of course considering you do want to proceed with further testing as well.

Official Channels

Please report any queries via security@bluemachines.ai providing all relevant information. We’ll do our best to reply within 5 business days starting from the reporting date. If we fail to get back to you after this time you may send another reminder after a week from the reporting date.

Our Commitments

  • Respond to your report promptly, and work with you to understand and validate your report.
  • Strive to keep you informed about the progress of a vulnerability as it is processed.
  • Work to remediate discovered vulnerabilities in a timely manner, within our operational constraints.
  • Extend Safe Harbor for your vulnerability research by not pursuing legal action as long as you comply with the terms in this document.
  • Not undermine your findings or behave in any unfair manner to avoid rewarding. For instance, calling a finding intended functionality even though it is not.

Our Expectations

This policy is designed to be compatible with common vulnerability disclosure good practices. It does not give you permission to act in any manner that is inconsistent with the applicable law, or which might cause us to be in breach of any legal obligations. In participating in our vulnerability disclosure program in good faith, we ask that you:

  • Play and follow the applicable law, including following this policy and any other relevant agreements. If there is any inconsistency between this policy and any other applicable terms, the terms of this policy will prevail.
  • Report any vulnerability you’ve discovered promptly. Do not go into further testing if you find any critical vulnerability that results in data loss/exfiltration, or Remote code execution of any sort, disruption of business continuity, system outage or jeopardizes the company’s reputation.
  • If you find any server-side vulnerability refrain from exploiting it in a way that may cause harm to user experience or expose our systems in unintended way (like exploiting a File upload vulnerability by uploading a full-blown payload/malware to our backend systems).
  • If you get any sort of CLI access immediately stop and inform us.
  • Avoid violating the privacy of others, disrupting our systems, destroying data, and/or harming user experience.
  • Use only the Official Channels to discuss vulnerability information with us. Do not send reports to any other address than the one stated in Official Channels (especially not an address that doesn’t have @bluemachines.ai in it). If any report is leaked it may result in legal actions against the researcher.
  • Use custom headers when possible which can help identify and isolate traffic, this can help expedite the triaging of vulnerability reports.
  • Provide us a reasonable amount of time (at least 30 days from the initial report) to resolve the issue before you ask for a public disclosure. The amount of time will depend on how critical/difficult the vulnerability is to remediate.
  • Perform testing only on in-scope systems, and respect systems and activities which are out-of-scope.
  • For POC purposes always use a custom payload that you created/have full knowledge of and which will not cause any sort of harm to our systems, environment, users, data or anything related to Blue Machines (like 3rd party vendors).
  • If a vulnerability provides unintended access to data: Limit the amount of data you access to the minimum required for effectively demonstrating a Proof of Concept; and cease testing and submit a report immediately if you encounter any user data during testing, such as Personally Identifiable Information (PII), Personal Healthcare Information (PHI), credit card data, or proprietary information
  • You should only interact with test accounts you own or with explicit written permission from the account holder
  • Do not engage in extortion.
  • You shall not exploit a security issue you discover for any reason other than for testing purposes, and you do not conduct testing outside your own account, a test account or another account for which you have the explicit written consent of the account owner to test. (This includes demonstrating additional risk, such as the risk that the security issue could be used to compromise sensitive company data or another user's account)
  • Do not access unnecessary, excessive, or significant amounts of data or modify data in our systems or services.
  • Do not disrupt our services or systems, use high‐intensity invasive or destructive testing methods.
  • If you inadvertently access another person's data or Blue Machines company data without authorization while investigating an issue, you must promptly cease any activity that might result in further access of user or Blue Machines data and notify Blue Machines what information was accessed (including a full description of the contents of the information) and then immediately delete the information from your system. Continuing to access another person's data or company data may demonstrate a lack of good faith and disqualify you from any benefit of the Safe Harbor provisions described below. You must also acknowledge the inadvertent access in any related bug bounty report you may subsequently submit. You may not share the inadvertently accessed information with anyone else.

Confidentiality

We will maintain confidentiality and exclusivity in the disclosure and remediation process. Likewise, you shall also maintain confidentiality and shall handle information including but not limited to description of vulnerability, shared findings, report, etc. with strict confidentiality. You shall not disclose any related information to third parties without written permission from our team.

Safe Harbor

When conducting vulnerability research, according to this policy, we consider this research conducted under this policy to be:

  • Authorized concerning any applicable anti-hacking laws, and we will not initiate or support legal action against you for accidental, good-faith violations of this policy.
  • Authorized concerning any relevant anti-circumvention laws, and we will not bring a claim against you for circumvention of technology controls.
  • Exempt from restrictions in our Terms of Service (TOS) and/or Acceptable Usage Policy (AUP) that would interfere with conducting security research, and we waive those restrictions on a limited basis and
  • Lawful, helpful to the overall security of bluemachines.ai, and conducted in good faith.
You are expected, as always, to comply with all applicable laws. If legal action is initiated by a third party against you, and you have complied with this policy, we will take steps to make it known that your actions were conducted in compliance with this policy in our sole discretion. However, Blue Machines cannot authorize any activity on third-party products without their written approval or guarantee they will not pursue legal action against you. We are not, in any way, responsible for your liability from actions performed on third parties.

Bug Bounty/Rewards

If your finding(s) is/are in accordance with all the rules stated above in Accepted Vulnerabilities section and does not violate any of our expectationsthen we'll be extremely thankful for your hard-work and time and reward you by showcasing your name and contribution in ourHall of Fame. If you have any blogs/want to publicly disclose the finding for educational purposes of the security community then we also will allow that once we fix the issue and notify you. In case we don't you can simply ask!

We understand it can be a little disheartening but Blue Machines does not have a bounty/cash reward for vulnerability disclosures as of now. We definitely do have plans for one in the future!

Please note: We only provide Hall of Fame mention to the first reporter of the vulnerability. Duplicates, internally known issues, or informative / OFS reports will not be eligible.

If at any time you have concerns or are uncertain whether your security research is consistent with this policy, please submit a report through one of our Official Channels before going any further. We understand that restricting your further testing will prevent us from understanding the full-blown impact of the vulnerability and will do our best to permit you for further testing.