HLD Group
Responsible disclosure
Guidelines for good-faith security research on HLD Group systems.
Last updated: 30 September 2026
Version 2.0 · Review cycle: 365 days · View all frameworks
1. Purpose and our commitment
HLD Group welcomes responsible, good-faith security research that helps us protect our customers, personnel, services, and the wider community. This policy explains which systems may be tested, how research must be conducted, what must be included in a report, how we will respond, and the conditions under which our safe-harbour commitment applies.
Security research can create real public benefit, but it can also create operational, privacy, and legal risk if it is performed without care. Researchers must minimise harm, use the least intrusive method reasonably available, keep information secure, stop when continued testing could cause harm, and give HLD Group a reasonable opportunity to investigate and remediate a finding before public disclosure.
We commit to treating researchers respectfully, assessing reports on their technical merits, communicating in good faith, and not pursuing or supporting legal action for research that is conducted in accordance with this policy. This is a coordinated vulnerability disclosure programme, not an invitation to access data, disrupt services, test unrelated third parties, or perform unrestricted penetration testing.
2. Who this policy applies to
This policy applies to independent security researchers, customers, partners, academics, and members of the public who discover a suspected security weakness in an HLD Group asset. HLD Group personnel and contractors must also report suspected vulnerabilities, but must use their internal security and incident-reporting channels where those channels are available.
By beginning or continuing testing under this policy, you confirm that you have read and will follow it, that you are acting in good faith, and that you will comply with applicable law. If you are testing for an employer, client, university, or other organisation, you must have authority to act for that organisation and must ensure it also understands any duties that arise from the research.
Minors and supervised researchers
Researchers who are not legally able to agree to these terms in their jurisdiction should ask a parent, guardian, teacher, or other authorised adult to submit the report and communicate with us. We will assess the technical report without discrimination, but recognition, confidentiality arrangements, or any discretionary reward may require the involvement of an adult who can lawfully accept them.
Sanctions and legal eligibility
We cannot make a payment, provide a benefit, or enter into an arrangement where doing so would breach applicable sanctions, export controls, anti-bribery laws, or other legal requirements. This restriction does not prevent a researcher from making a safety report. Where lawful, we will still receive and assess the technical information.
3. Definitions
For this policy, an “asset” is a website, API, application, service, repository, system, or device operated by HLD Group. “Customer data” includes any information submitted by, generated for, or relating to a customer or its users. “Personal data” means information relating to an identified or identifiable person. “Testing” includes manual or automated activity intended to identify, confirm, or demonstrate a security weakness.
A “vulnerability” is a weakness that could reasonably compromise confidentiality, integrity, availability, authentication, authorisation, privacy, or another security control. “Good faith” means honest research intended to improve security, performed carefully and without extortion, coercion, deception, competitive misuse, personal benefit from accessed data, or intent to harm. “Public disclosure” includes publishing details, demonstrations, screenshots, exploit code, indicators, or hints that would enable another person to identify or exploit the issue.
4. Assets in scope
Only the assets listed below are in scope, and only to the extent that HLD Group operates and controls them. An HLD Group name, logo, link, DNS record, integration, or data flow does not by itself establish that an asset is operated by HLD Group. If ownership is unclear, ask us before testing.
- Public web properties on hldgroup.org and its subdomains that are operated by HLD Group
- Publicly documented HLD Group APIs and the authentication, authorisation, and application logic exposed through them
- HLD Group applications and services made generally available to customers, where testing is performed only through accounts and tenants that you own or are expressly authorised to use
- Customer trial, demonstration, or sandbox environments that are expressly labelled as available for security testing
- Open-source repositories in an official HLD Group organisation, limited to vulnerabilities in code maintained by HLD Group and not unrelated weaknesses in the hosting platform
- HLD Group mobile or desktop applications obtained through an official distribution channel, excluding the device, operating system, application store, and third-party libraries except where our implementation creates the vulnerability
Customer environments
A customer environment is in scope only if you are the customer, are an authorised user of that customer, or have the customer’s explicit written permission. This policy does not authorise testing another customer’s tenant, users, configuration, data, integrations, or endpoints. Evidence of cross-tenant exposure should be demonstrated using two accounts or tenants that you control wherever possible.
Recently acquired, discontinued, or legacy assets
Assets of an acquired business, discontinued product, legacy brand, parked domain, or former service are not automatically in scope. You may report a suspected issue, but you must not actively test the asset unless HLD Group confirms in writing that it is controlled by us and covered by this policy.
5. Assets and activities out of scope
The following are out of scope unless HLD Group gives you specific written authorisation in advance. A report about an out-of-scope issue may still be useful, but the safe-harbour commitment in this policy does not authorise the underlying testing.
- Systems, applications, networks, accounts, offices, devices, or data belonging to customers, suppliers, partners, personnel, or any other third party
- Third-party platforms used by HLD Group, including cloud, identity, payment, analytics, content-delivery, support, communications, application-store, source-hosting, and social-media providers
- Personal accounts, personal devices, home networks, or personal email addresses of HLD Group personnel or contractors
- Production infrastructure that is not publicly exposed for ordinary use, including internal administrative systems and management interfaces
- Any asset for which a separate vulnerability disclosure or bug-bounty policy applies, unless that policy expressly refers back to this one
- Denial-of-service, distributed denial-of-service, resource exhaustion, destructive testing, traffic flooding, or activity likely to impair availability or materially increase cost
- Social engineering, phishing, vishing, smishing, pretexting, impersonation, bribery, threats, harassment, or attempts to obtain credentials or access from a person
- Physical intrusion, tailgating, surveillance of personnel, hardware tampering, facility testing, mail-based attacks, or attacks against safety systems
- Malware deployment, ransomware, cryptomining, botnets, persistent access, command-and-control activity, or establishment of backdoors
- Accessing, altering, downloading, retaining, transmitting, or deleting data that does not belong to you, except for the minimal accidental access addressed in section 10
- Testing from infrastructure that conceals abusive activity, uses compromised hosts, or prevents us from contacting or safely blocking the testing source
6. Research rules and standard of care
You must conduct research in a manner that is proportionate to the suspected risk. Start with observation and non-invasive validation. Escalate only as far as necessary to establish that a vulnerability exists and to explain its likely impact. A report does not need to prove the maximum conceivable impact, and you must not create additional victims or damage merely to strengthen a severity claim.
- Use accounts, organisations, tenants, content, payment methods, and test data that you own or are expressly authorised to use
- Clearly mark test accounts and data where practical, and remove test content when the research is complete if it is safe to do so
- Keep request rates low, avoid peak periods where known, and stop if you observe instability, elevated latency, errors, alerts, or unexpected resource consumption
- Use the minimum number of requests, records, privileges, and systems necessary to validate the issue
- Do not pivot from the affected component into other systems, seek persistence, evade monitoring, erase logs, or conceal the nature of your activity
- Do not exploit a vulnerability for profit, competitive intelligence, surveillance, retaliation, publicity leverage, or access to services you have not paid for
- Comply promptly with a reasonable request from HLD Group to pause or stop testing while we investigate a safety or operational concern
- Preserve an accurate timeline, relevant request identifiers, tool settings, source IP addresses, and other evidence that can help us distinguish your research from malicious activity
7. Permitted testing techniques
The following techniques are generally permitted when applied carefully to in-scope assets and in accordance with every other requirement in this policy. Permission is limited to validation; it is not permission to cause material impact or obtain data belonging to others.
Application and API testing
- Testing authentication, session management, authorisation, object-level access control, input validation, output encoding, request forgery protections, file handling, business logic, and security headers
- Using two or more accounts that you control to demonstrate horizontal or vertical privilege issues without involving another user
- Submitting benign payloads that demonstrate injection or code-path control without executing destructive commands, reading unrelated data, or modifying the underlying service
- Inspecting client-side code, application packages, documented endpoints, and network traffic generated by your own authorised use
Automated tools
Low-volume automated scanning is permitted only where it does not degrade the service, generate excessive traffic, create large volumes of records, or target third parties. You remain responsible for the tool’s configuration and actions. Scanner output without manual validation and a clear security impact is normally insufficient for triage.
Open-source and supply-chain research
You may analyse HLD Group-maintained source code and report vulnerable code paths, exposed secrets, unsafe build or release behaviour, and dependency risks that have a demonstrable effect on an in-scope asset. Do not publish or use a discovered credential. Do not test a dependency maintainer’s infrastructure under this policy; report the upstream issue to the party that controls it.
8. Prohibited testing techniques
The following techniques are prohibited because they create disproportionate risk or do not require live validation. If you believe one is necessary to demonstrate an exceptional issue, contact us with a testing plan and wait for specific written authorisation.
- Password spraying, credential stuffing, brute force, multi-factor authentication fatigue, or testing credentials obtained from breaches or illicit sources
- Creating excessive accounts, organisations, transactions, support requests, notifications, emails, SMS messages, files, logs, or other records
- Sending unsolicited messages or payloads to customers, personnel, or third parties, even as a proof of concept
- Executing commands that alter or destroy data, changing account ownership, disabling security controls, rotating another user’s credentials, or locking users out
- Extracting database contents, cloud metadata, secrets, tokens, private keys, source code, backups, internal documents, or personal data beyond the minimum evidence permitted by section 10
- Installing software or certificates on a system you do not own, maintaining access after validation, or attempting to evade detection or attribution
- Testing live payment instruments, initiating chargebacks, committing fraud, obtaining goods or services without payment, or manipulating financial, billing, credit, payroll, or tax records
- Uploading unlawful, harmful, intimate, classified, export-controlled, or genuinely malicious content as test data
- Publicly disclosing an unremediated vulnerability, selling or transferring it to another party, or threatening disclosure to obtain payment or another advantage
9. Authentication, accounts, and identity scenarios
Account takeover and reset flows
Test takeover scenarios only against accounts you control. Do not attempt to reset, recover, enumerate, or assume control of another person’s account. Where a flaw appears to expose whether an account exists, use addresses and identifiers you own and explain how the behaviour could affect others without collecting a list of users.
Leaked credentials or secrets
If you find an HLD Group credential, token, private key, or secret in a public location, do not use it to sign in or determine its privileges. Report the location and enough of the value to identify it, with the remainder redacted. If a secret appears in a private or illicit dataset, do not send us the dataset; describe the source and follow applicable law.
Multi-factor authentication and identity providers
You may test enrolment, recovery, and session behaviour on your own accounts. Do not trigger prompts, messages, calls, or recovery workflows for another person. Identity-provider weaknesses that are independent of HLD Group configuration must be reported to the provider under its own policy.
10. Data access, privacy, and accidental exposure
Do not intentionally access data belonging to another person or organisation. If data appears unexpectedly, stop immediately. Do not continue browsing, search for additional records, change permissions, download a collection, take unnecessary screenshots, or share the data with anyone other than HLD Group through the reporting channel.
Capture only the minimum evidence needed to identify the affected endpoint and nature of the exposure. Prefer record counts, field names, synthetic examples, hashes, redacted screenshots, or a description over actual content. If one record is sufficient to demonstrate the issue, do not access a second record. Do not retain special-category, health, financial, identity, authentication, communications, child-related, or other highly sensitive data unless we specifically ask you to preserve a minimal sample and doing so is lawful.
- Tell us what type of data was exposed, approximately how much was visible, when it occurred, and whether you believe anyone else accessed it
- Secure any evidence at rest and in transit, restrict access to people directly involved in the report, and do not place it in public issue trackers or consumer file-sharing links
- Delete HLD Group and customer data as soon as we confirm it is no longer needed, and confirm deletion if we request it
- If you accidentally changed or deleted data, tell us exactly what changed; do not attempt an uncoordinated repair that could worsen the impact
- If the exposure presents an immediate risk to people, state “URGENT — DATA EXPOSURE” in the email subject so that we can prioritise containment
11. Availability, performance, and denial of service
Do not test whether a service can be overwhelmed. This includes distributed or single-source flooding, large uploads, decompression bombs, high-cost queries, mass export or report generation, queue saturation, cache exhaustion, deliberate crash loops, and repeated actions that create material infrastructure or third-party charges.
A denial-of-service weakness can usually be reported from code analysis, a single benign request, a locally reproduced test, or a reasoned explanation. If you unexpectedly affect availability or performance, stop all related traffic, preserve your logs, and report the event immediately. Do not repeat the test to confirm that the impact was yours.
12. Social engineering, physical security, and people
Do not deceive, pressure, monitor, or target HLD Group personnel, customers, contractors, support agents, suppliers, or their families. This includes phishing simulations, impersonation, pretext calls, malicious documents, QR codes, gift or payment offers, consent phishing, support-channel manipulation, and attempts to bypass a control through a person.
Do not enter or surveil offices, data centres, residences, or restricted areas; follow personnel; inspect unattended devices; search rubbish; connect rogue hardware; or send physical media. You may report an accidentally observed physical or procedural weakness without attempting to exploit it.
13. Third parties, integrations, and shared responsibility
Many HLD Group services interoperate with third-party products. Our policy cannot authorise testing of a third party. Limit testing to the HLD Group side of an integration and to accounts you control on both sides. Follow the third party’s policy for any testing that reaches its systems.
If the root cause is in a third-party product but HLD Group’s configuration or use creates a distinct exposure, report it to us and explain the boundary. If the same weakness affects other organisations, do not test those organisations. We may coordinate with the supplier, an industry coordinator, or a computer emergency response team, and we may ask you to participate in that process.
14. Artificial intelligence and automated decision systems
Security issues involving AI features are in scope when they create a concrete impact on an HLD Group security boundary. Examples may include unauthorised access to another user’s data, privilege escalation through tool use, cross-tenant retrieval, unintended execution of privileged actions, or bypass of an access control. Model behaviour that is merely inaccurate, subjective, undesirable, or reproducible only through ordinary prompting without a security impact is not normally a vulnerability.
- Use only synthetic or non-sensitive prompts and content that you are legally entitled to submit
- Do not attempt to extract another user’s prompts, files, memories, credentials, personal data, system instructions, proprietary datasets, or model weights
- Do not connect an AI feature to third-party systems or grant it permissions for the purpose of testing unless you own and control those systems and permissions
- Do not cause an agent or automated feature to contact people, make purchases, change production data, publish content, execute malware, or take another consequential external action
- Clearly distinguish a confirmed security-boundary failure from a hypothetical prompt, content-safety concern, quality issue, or model hallucination
15. What to report
Send a report as soon as reasonably possible after discovery. One well-structured report per distinct root cause helps us route and remediate issues efficiently. Closely related impacts from the same root cause may be combined. If you are unsure whether findings share a root cause, explain the relationship rather than submitting a large volume of near-identical reports.
- Your name or preferred handle, an email address we can use, and whether you want public recognition if the report is validated
- The affected hostname, application, API route, repository, component, version, account type, and environment
- A concise vulnerability summary and the security property that fails
- Numbered, reproducible steps using the least destructive proof of concept available
- Expected behaviour, observed behaviour, prerequisites, required privileges, and whether user interaction is necessary
- A realistic impact analysis describing who could exploit the issue, who could be affected, and what data or capability is at risk
- Dates and times with a time zone, source IP addresses used for testing, relevant request or correlation identifiers, and user-agent or tool details
- Sanitised request and response samples, screenshots, logs, or short video where useful, with credentials, tokens, personal data, and customer information redacted
- Any actions already taken, data accidentally accessed, service impact observed, public disclosure already made, or other party already notified
- Suggested remediation or references, if available; a proposed fix is helpful but is not required
Reporting channel
Email security@hldgroup.org with a descriptive subject line. For a critical issue, active exploitation, exposed secret, material data exposure, or ongoing service impact, begin the subject with “URGENT”. Do not include live credentials, unnecessary personal data, or executable malware in the first message. If the report contains sensitive material, ask for an encrypted exchange or use our published PGP key where available.
Attachments and proof-of-concept code
Prefer plain text, PDF, common image formats, or a private repository made available to our security team. Describe archives and unusual file types before sending them. Proof-of-concept code must be narrowly scoped, must not self-propagate or persist, and must not contain real secrets or data. We may decline to open unexpected binaries, password-protected archives without prior coordination, or links that require unknown browser extensions or executables.
16. Reports that are normally not actionable
We review reports in context. The following observations are commonly low-risk, informational, intended behaviour, or dependent on a demonstrable impact. We may close them without remediation unless the report shows a credible security consequence specific to our implementation.
- Missing security headers or cookie attributes without an exploitable impact, including purely best-practice reports
- TLS configuration preferences, supported cipher observations, or DNS and email-security records without a practical attack
- Version disclosure, banners, stack traces, directory paths, or publicly available metadata that do not expose sensitive information or enable exploitation
- Clickjacking on pages that perform no sensitive action, self-XSS, text injection, or issues requiring a victim to paste attacker-controlled code into developer tools
- Open redirects, tabnabbing, CSV formula injection, content spoofing, or rate-limit observations without a credible and material security impact
- Username or email enumeration that does not materially increase risk, especially where the information is already public or required by the product flow
- Logout, session duration, password-complexity, or multi-factor authentication preferences that reflect documented product or customer configuration
- Scanner-only results, dependency version matches, CVE references, or source-code patterns without evidence that the vulnerable path is present, reachable, and exploitable
- Hypothetical subdomain takeover without control of the referenced external resource, or claims based only on a missing or unclaimed DNS record
- Email spoofing claims without delivery evidence and alignment analysis, spam, unsolicited bulk email, or abuse that does not arise from a technical vulnerability
- Attacks requiring physical access to an already unlocked device, a rooted or jailbroken device, a compromised operating system, or full control of the victim’s browser
- Secrets, credentials, or personal information that are fabricated, expired, intentionally public, belong to the reporter, or originate solely from an unrelated third-party breach
- AI prompt injection, jailbreaks, hallucinations, bias, undesirable output, or content-safety bypasses that do not cross an HLD Group security or authorisation boundary
- A vulnerability that has already been reported, is already being remediated, or shares the same underlying cause as an earlier report
17. Triage, severity, and duplicate handling
We assess reports using technical severity, exploitability, required access and interaction, affected population, data sensitivity, operational impact, detectability, compensating controls, and the likelihood of real-world abuse. Industry scoring systems such as CVSS may inform the assessment but do not determine it by themselves. The severity assigned by HLD Group may differ from the reporter’s assessment.
A report is a duplicate when we already knew of the same underlying weakness, including through internal testing, another researcher, a supplier, or an incident. Different endpoints, payloads, or affected records do not necessarily make findings distinct if one remediation addresses them. We will tell you when we can that a report is a duplicate, but confidentiality and privacy obligations may prevent us from identifying the earlier source or sharing its details.
Informational and accepted-risk findings
A report may be valid but informational, mitigated by other controls, or accepted as a business risk. Validation does not guarantee a code change, public advisory, identifier, recognition, or reward. We may implement a different remediation from the one proposed where it provides an equivalent or stronger risk reduction.
Abusive or misleading submissions
We may restrict contact or reject reports that contain threats, discriminatory or abusive language, fabricated evidence, repeated demands, spam, plagiarism, malicious attachments, intentional overstatement, or pressure for payment. A disagreement about severity or eligibility does not justify disclosure of sensitive information or further exploitation.
18. Our response process and service targets
We aim to provide a fair and predictable process. The targets below are goals rather than guarantees and are measured in HLD Group business days. Complex, third-party, cross-border, or incident-related reports may take longer. A delay does not authorise additional testing or public disclosure where doing so would create a material risk.
- Acknowledgement within five business days, including a tracking reference where practicable
- Initial triage within ten business days, with a request for missing information or an initial view of validity where practicable
- Priority escalation for credible critical vulnerabilities, active exploitation, exposed production credentials, material data exposure, or ongoing availability impact
- Reasonable progress updates for validated findings, generally when status materially changes or at an interval agreed with the researcher
- Notification when a validated issue is remediated or otherwise resolved, where legal, security, contractual, and operational constraints permit
Researcher responsiveness
Please respond to reasonable questions and requests for clarification. If we cannot reproduce a finding and receive no response after reasonable follow-up, we may close the report. You may reply later with new information and ask us to reopen it.
Retesting
When we invite retesting, use the original proof of concept and remain within the scope we specify. Do not broaden the test to unrelated systems or data. A production deployment may be phased, and the continued presence of a symptom does not authorise escalation beyond the original validation steps.
19. Remediation and coordinated disclosure
Please allow us a reasonable opportunity to investigate, develop and test a fix, deploy it safely, notify affected parties where appropriate, and coordinate with suppliers or customers. As a general starting point, we ask for up to 90 days from our acknowledgement before public disclosure. The appropriate period may be shorter for a simple fix or longer for complex infrastructure, safety-sensitive changes, coordinated industry remediation, or vendor dependencies.
Do not publicly disclose technical details, exploit code, affected customer information, unpatched endpoints, or remediation bypasses until we confirm that coordinated disclosure can proceed or the agreed period expires. Before publication, share the proposed content and allow a reasonable review period so we can identify sensitive data, factual errors, or details that would unnecessarily expose users. We will not use review to suppress legitimate research or criticism.
- We may ask to accelerate disclosure when users need information to protect themselves or active exploitation is already public
- We may ask for additional time where a fix is unusually complex, a supplier controls remediation, or immediate disclosure would create a substantial and continuing risk
- We may publish an advisory, release note, CVE, acknowledgement, or no public notice, depending on impact and the needs of affected users
- Neither party should represent that the other endorses its publication, conclusions, product, or services without express permission
- Embargoed information must remain confidential until the agreed disclosure date or until HLD Group makes the information public
20. Safe harbour
If you make a good-faith effort to comply with this policy, HLD Group will consider your research to be authorised with respect to the HLD Group assets covered by this policy. We will not initiate or support legal action against you for accidental, proportionate violations that occur despite that good-faith effort, provided you promptly stop, disclose the circumstances, minimise harm, and continue to cooperate.
If a third party initiates legal action against you and your research complied with this policy, we may, at our discretion and where lawful, confirm that your activity was conducted under our responsible disclosure programme. We cannot bind prosecutors, regulators, courts, customers, suppliers, platform operators, or other third parties, and cannot authorise activity on systems they control.
Safe harbour does not apply to extortion, threats, unlawful acquisition or use of data, intentional harm, privacy invasion, physical or social-engineering attacks, denial of service, persistence, concealment, testing after a request to stop, public disclosure that creates unreasonable risk, or other activity materially outside this policy. Where you are uncertain whether planned testing is permitted, contact us first and wait for written clarification.
No waiver of legal protections
This policy does not require you to waive rights or protections available under applicable law, and it does not create a duty to perform research. It does not grant access rights, intellectual-property licences, confidentiality clearances, employment, agency, partnership, or authority to act on behalf of HLD Group beyond the limited testing permission expressly described here.
21. Recognition, rewards, and expenses
HLD Group does not currently operate a standing paid bug-bounty programme. Submitting a report does not create an entitlement to payment, reimbursement, employment, a contract, a gift, or any other benefit. Do not incur costs in the expectation that we will reimburse them.
We may, in our sole discretion and subject to law, offer recognition or a reward for a particularly significant and well-reported finding. Any offer will be made explicitly in writing and may be subject to identity, eligibility, tax, sanctions, and payment checks. The existence or amount of a discretionary reward for one report does not establish a precedent for another.
With your consent, we may acknowledge a validated report in our security researcher hall of fame using your name, handle, affiliation, and link. We will not knowingly publish your identity without permission. Recognition may be withheld for duplicate, informational, low-quality, out-of-scope, unlawfully obtained, or policy-violating reports.
22. Confidentiality, intellectual property, and report handling
Treat non-public information learned through research or our response as confidential until it is lawfully public or we agree otherwise. You retain ownership of your original report and tools. By submitting a report, you grant HLD Group a worldwide, non-exclusive, royalty-free right to receive, reproduce, analyse, test, adapt, remediate, and share the report and proof of concept as reasonably necessary to investigate and resolve the issue, protect users, meet legal obligations, and coordinate with affected suppliers or advisers.
Do not submit information that you are not authorised to disclose, including another organisation’s confidential information, unlawfully obtained datasets, or material subject to restrictions that prevent us from using it. Tell us before submission if you believe special handling terms are necessary. We are not bound by terms included unilaterally in a report, signature, portal, or attachment unless an authorised HLD Group representative agrees to them in writing.
We protect report information according to its sensitivity and retain it for security, audit, legal, and programme-integrity purposes. We may share it on a need-to-know basis with personnel, professional advisers, insurers, hosting providers, affected customers, suppliers, regulators, law enforcement, or vulnerability coordinators where appropriate and lawful.
23. Special situations and escalation
Active exploitation or immediate danger
If you believe exploitation is active, a production secret is publicly exposed, data is being removed, or people face immediate harm, stop testing and email security@hldgroup.org with “URGENT” at the start of the subject. Include concise containment information first. Do not attempt to investigate the attacker, take over infrastructure, notify affected individuals, or contact the media on our behalf.
You caused an incident
If research accidentally causes instability, data change, notification, cost, account impact, or another incident, stop immediately and report exactly what happened. Preserve local evidence and remain available. Prompt, candid reporting and cooperation are important evidence of good faith; concealment or repetition is not.
Publicly known or already disclosed vulnerabilities
A known CVE or public technique can still be relevant if you demonstrate that an in-scope HLD Group asset is affected. Do not exploit it beyond minimal validation. If active exploitation is widely reported, prioritise rapid private notice rather than public proof that identifies our exposure.
Law-enforcement or regulatory contact
Nothing in this policy prevents either party from complying with a lawful obligation. If you believe a disclosure must be made to a regulator, computer emergency response team, or law-enforcement body, tell us in advance where lawful so that evidence and user protection can be coordinated.
24. Practical examples
These examples illustrate how the policy applies. They do not override the requirements above, and an unlisted scenario is not automatically permitted.
Cross-tenant access
Permitted: create two tenants you control, show that a user in one can access a harmless synthetic record in the other, and stop. Not permitted: enumerate customer identifiers, open a real customer record, download a customer dataset, or change another organisation’s settings.
Injection and remote execution
Permitted: use a benign expression or command such as returning a fixed marker to demonstrate control. Not permitted: list directories, read environment variables, retrieve secrets, establish a shell, install a payload, pivot, or retain access.
Exposed storage or backups
Permitted: report the location, response metadata, and a redacted indication that access controls are missing. Not permitted: list all objects, download an archive, open customer files, or search the contents for sensitive information.
Rate limiting and enumeration
Permitted: make a small number of requests using identifiers you own and explain the abuse case. Not permitted: run a large wordlist, send messages to real users, test breached passwords, or continue until infrastructure or account protections activate.
Payment and commerce logic
Permitted: demonstrate an inconsistency using a sandbox, documented test instrument, or transaction that can be safely cancelled. Not permitted: retain goods or services without payment, use another person’s payment method, create a chargeback, alter a real invoice, or move funds.
Stored content and notifications
Permitted: store a harmless marker in your own private test area and view it through your own second account where needed. Not permitted: place payloads where customers or personnel may encounter them, send unsolicited email or SMS, or use alarming or deceptive content.
Secrets in source code
Permitted: identify the repository, file, commit, and a short redacted fingerprint. Not permitted: authenticate with the secret, determine what data it can access, copy the repository elsewhere, or publish the value.
AI tool use
Permitted: use synthetic data to show that an HLD Group AI feature can invoke a privileged tool across an authorisation boundary. Not permitted: ask the feature to expose another user’s content, contact a real person, execute destructive actions, or ingest confidential third-party material.
25. Other reports and support requests
The security channel is for technical vulnerabilities and urgent security exposure. Product support questions, account recovery, billing disputes, suspected fraud without a product vulnerability, privacy-rights requests, complaints, copyright notices, sales enquiries, and general abuse reports should use the appropriate published contact channel. Sending those matters to security may delay the response.
If you are unsure whether an observation is a vulnerability, you may still contact us. Describe what you observed and avoid further intrusive testing while you wait. We would rather receive a concise good-faith question than have a researcher take unnecessary risk to prove impact.
26. Policy administration and changes
HLD Group’s security function owns this policy. We review it periodically and may update it as our services, threat environment, or legal obligations change. The version published when research occurs will generally guide our assessment of that activity, although a later version may apply where it offers clearer or more favourable protection to a good-faith researcher.
Material changes will be reflected by the “Last updated” date on this page. No employee or contractor may vary this policy or authorise additional testing unless they have authority to do so and provide the variation in writing. Questions about scope, proposed testing, report status, or this policy should be sent to security@hldgroup.org.
For contractual attestations or audit packs, contact security@hldgroup.org.