What SendSafe replaces
EU direct identifiers: name, BSN, IBAN, email address, phone number, city. Detection lowers the risk but is never complete.
Replaced on sendExample
- Name
- PERSON_1
- EMAIL_1
- Phone
- PHONE_1
- IBAN
- ACCOUNT_1
- City
- CITY_1
Security
Security and trust, for the people who sign off.
For IT, security and privacy leads. SendSafe runs on the device and swaps personal data for neutral labels before a request leaves for an AI tool. This page sets out what stays on the device, what it sends, and where its responsibility ends.
SendSafeProtecting
ChatGPTPrompt sent
- Anna de VriesPERSON_1
- anna@devries.exampleEMAIL_1
- +31 6 1420 2287PHONE_1
- NL91 ABNA 0417 1643 00ACCOUNT_1
- UtrechtCITY_1
The mapping stays on this device
What leaves the device, and what doesn't.
Originals stay on the device. Only neutral labels go to the AI tool.
Stays on your device
The sensitive parts
- Original valuesThe values SendSafe recognised and replaced.
- The mappingWhich label stands for which value.
- The restored answerAssembled on the device when the response returns, on Mac and Windows.
Sent to the AI provider
What the AI needs to help
- The question and its context"Draft a short reply to PERSON_1 …"
- LabelsNeutral labels such as EMAIL_1, sent in place of the values behind them.
- Anything it did not recogniseA value detection misses goes along as typed.
Originalstays local
Check the refund to NL91 ABNA 0417 1643 00.Sentleaves the device
Check the refund to ACCOUNT_1.Mappingstays local
ACCOUNT_1NL91 ABNA 0417 1643 00Scope, policy and providers.
The details an IT team asks about first.
Rules and reports per team
IT chooses which categories stay local for each team, once. Reporting follows the same line: counts by team, category and tool, never by person and never with content.
- Name
- BSN
- IBAN
- Email address
- Phone number
- City
Switch models, keep your policy.
The same local protection sits in front of each of the major providers. Adding or changing a model does not mean rewriting policy.
+4What IT sees.
The admin dashboard shows coverage, categories, AI tools and policy per team. Prompt content never reaches it, and usage is never tied to a person.
For your security review.
What we can confirm today. Anything else, ask us in the walkthrough.
Found a vulnerability? Email security@sendsafe.ai.
- Where values are replaced
- On the device, when you send. Neutral labels go to the AI provider, the originals stay on the device.
- What is replaced
- EU direct identifiers: name, BSN, IBAN, email address, phone number, city.
- Prompt archive
- None. SendSafe does not build a prompt archive.
- Reporting
- Counts per team, category and AI tool. Never per person and never with content.
- Platforms
- macOS, Windows, iOS and Android. The answer is restored on the device on macOS and Windows.
- AI providers
- Your organisation keeps its own accounts and contracts with the AI providers it approves.
- Legal entity
- OneMinded B.V., High Tech Campus 5, Eindhoven. Chamber of Commerce 97404829.
- Data processing agreement (DPA)
- We sign a data processing agreement with every customer before the pilot starts. Request it via privacy@sendsafe.ai.
Subprocessors of this website
- Vercel Inc.United States
- Hosting, server logs and cookieless visitor statistics: IP address, user agent, pages visited. Covered by appropriate safeguards, such as the EU Standard Contractual Clauses or the EU-US Data Privacy Framework.
- Upstash Inc.United States
- We store form submissions with Upstash, in their EU region (Frankfurt). Upstash is a US company. Any access from the US is covered by the EU Standard Contractual Clauses or the EU-US Data Privacy Framework.
The product's subprocessors are listed in the data processing agreement.
A technical measure, not a full compliance programme.
SendSafe can reduce what reaches an AI provider. It does not replace the decisions your organisation still has to make.
What stays with your organisation
The decisions around AI use
- Lawful basis and policiesWhich AI use is allowed, for what, and by whom.
- Provider contractsData processing terms with the AI providers you approve.
- Assessments and recordsDPIAs, risk assessments and documentation where required.
Certification status. SendSafe does not currently claim any security certification or attestation. If that changes, it will be stated here with the scope and date. Questions for a security review are welcome at security@sendsafe.ai.
Bring your security questions.
We walk IT and security teams through the architecture, and scope a pilot around your requirements.