Responsible disclosure
If you discover a vulnerability in our systems, we would like to hear about it. We take every report seriously, we keep you informed, and we take no legal action against reporters who follow the rules on this page.
security@flowstead.aiAt Flowstead we build digital colleagues that carry out work independently in the back office of manufacturing companies. Our customers trust us with their orders, invoices, contracts and customer data. We take the security of that data seriously, and we also know that no system is flawless.
This policy follows the Coordinated Vulnerability Disclosure guideline of the Dutch National Cyber Security Centre.
How to report a vulnerability
Send your report to security@flowstead.ai. Preferably write in Dutch or English.
Help us assess your finding quickly by including at least the following:
- A description of the vulnerability and the system, domain or endpoint where you found it.
- The steps we need to reproduce the problem, ideally detailed enough for us to recreate it ourselves.
- The risk you see, in other words what an attacker could achieve with this vulnerability.
- Any screenshots, log lines or scripts you used.
- How we can reach you with follow-up questions.
If you want to communicate confidentially or encrypted, say so in your first message. We will then contact you to agree on a secure channel.
What is in scope
This policy applies to systems that belong to Flowstead and that we control:
- flowstead.ai and all its subdomains.
- The Flowstead platform, including the environment where digital colleagues are configured and managed.
- Our APIs, including the interfaces through which customer systems communicate with the platform.
Vulnerabilities in digital colleagues
Because our digital colleagues run on language models, we explicitly count the following findings as in scope:
- Prompt injection that pushes a digital colleague into actions outside its assignment, for example through the content of an incoming email, an invoice or an attachment.
- Breaking the separation between customer environments, where data or instructions from one customer become visible or usable in another customer's environment.
- Bypassing the approval steps that determine which actions a digital colleague may take on its own and which a human has to confirm.
- Unintended exposure of data through model output, log files, error messages or export functions.
If you are unsure whether your finding is in scope, report it anyway. We would rather judge that ourselves than miss something.
What is out of scope
We do not process the following reports:
- Vulnerabilities in third-party systems and services. We will point you to the party involved and are happy to help make that contact.
- Output of automated scans without demonstrated impact.
- Missing or suboptimal security headers, without concrete demonstrable abuse.
- Missing best practices in SSL and TLS configuration without demonstrable risk.
- Self-XSS and clickjacking on pages without authenticated actions.
- Vulnerabilities that can only be exploited with physical access to a device, or with an already compromised account or device.
- Reports that concern only the substantive quality or accuracy of model output. A digital colleague that makes a mistake is a quality issue for us, not a security incident, unless a third party can force that mistake.
- Denial of service, brute force, and other tests that affect the availability of our service.
- Social engineering aimed at our employees, customers or suppliers.
- Physical attacks on our offices or equipment.
- Reports about the contact form or the newsletter signup on our website.
Rules for reporters
We ask you to follow these rules during your research.
Do no more than needed to demonstrate the problem
Once you can show that you have access to data, stop there. Do not view, copy, change or delete data that is not yours.
Do no damage
Do not change or delete system settings, do not place a backdoor, and do not alter how our service works, not even to prove that you could.
Do not share your finding with others
Keep it between us until we have fixed the problem. We publish by mutual agreement, see the section on timelines.
Report as soon as possible
Report shortly after you find the vulnerability, so that we can act quickly.
Do not affect the availability of our service
Do not use denial of service, brute force on login pages, spam, or large-scale automated querying of our systems.
No social engineering or physical access
Do not use phishing, deception of employees or physical access to get in.
Stay out of customer environments
Do not conduct research in our customers' production environments or on the connections to their systems. Those systems are not ours and fall outside this policy.
What you can expect from us
- We respond to your report within three working days with an initial assessment and, where possible, an expected fix date.
- We keep you informed of progress until the vulnerability has been resolved.
- We treat your report confidentially and do not share your personal data with third parties without your consent, unless we are legally required to do so.
- We will not report you to the authorities and will take no legal action against you, provided you follow the rules in this policy and your report was demonstrably aimed at improving our security.
- We will credit you by name in our publication about the resolved vulnerability if you want us to. Tell us how you would like to be named, or let us know that you prefer to stay anonymous.
We cannot rule out that the Dutch Public Prosecution Service decides to prosecute on its own initiative, but in weighing that decision it does take into account whether a reporter has evidently followed a CVD policy.
Timelines
First response
3 working days
With an initial assessment and, where possible, an expected fix date.
Target fix time
60 days
The period the Dutch NCSC uses as a guideline for software.
Publication
By agreement
In principle only after the problem is fixed, agreed together with you.
We aim to resolve a reported vulnerability within 60 days. If a problem is more complex, we will say so and agree a realistic timeline with you.
We publish about the vulnerability in consultation with you and in principle only after the problem is fixed. If fixing it takes longer, we agree together on a reasonable moment to publish. We use 60 days after the report as the starting point for that conversation.
Rewards
Flowstead has no bug bounty programme and pays no monetary rewards for reports. What we do instead: we thank you personally, we credit you by name in the publication if you appreciate that, and we take the time to walk you through what we did with your finding.
Not sure whether your finding belongs here?
If you have a question about this policy, just email us. We are happy to think along with you.
security@flowstead.aiThis policy was last changed on 31 August 2026. Version 1.0.
Flowstead, registered in Utrecht, the Netherlands.