Is SIGNLD secure

In this article

• How does SIGNLD encrypt data
• Can SIGNLD write back to my systems
• Where does the AI inference run
• Does SIGNLD use third-party AI APIs
• Is my data used to train the model
• Is every query tracked
• How does the Knowledge Graph handle sensitive data
• Who can see a Decision Brief
• What SIGNLD does not claim

How does SIGNLD encrypt data

Data at rest is encrypted with AES-256. Data in transit between connected systems, the Knowledge Graph, and the application uses TLS 1.2 or higher. These two encryption standards apply to every connection and every Topic, regardless of account tier.

For the wider context, see our overview of what SIGNLD is.

Encryption is enforced at the infrastructure level, not configured per customer. A company connecting its first system and a company running many Topics get the identical encryption standard.

There is no setting to turn encryption off or downgrade it for a specific system, department, or Topic. The standard is fixed across the entire platform rather than being a configurable option a customer could disable.

Can SIGNLD write back to my systems

No. Every connection SIGNLD makes to a source system, whether a CRM, an accounting platform, or a job management tool, is read-only. SIGNLD reads records to build the Knowledge Graph and answer questions. It does not create, edit, or delete records in any connected system.

This read-only design means a misconfigured Topic or a bad question cannot alter data in the systems your team already relies on. The connection can only be used to pull information out, never push information back in.

Because the connection has no write permission at the credential level, there is no code path inside SIGNLD that could accidentally trigger a write to a connected system. The restriction sits at the access layer, not just in application logic.

Where does the AI inference run

AI inference runs on a single-tenant AWS Bedrock instance. This is a private large language model deployment powered by AWS Bedrock, dedicated to a single customer rather than shared across multiple organizations. Requests from other SIGNLD customers do not pass through the same instance.

The single-tenant model separates inference infrastructure by customer account. This is distinct from a multi-tenant setup where different customers' requests are processed on shared compute.

This separation holds regardless of account tier or the number of Topics a customer runs. A customer on the free tier and a customer on Enterprise both get a dedicated inference instance rather than a shared pool.

Does SIGNLD use third-party AI APIs

No. SIGNLD does not send data to third-party AI APIs. All AI processing happens within the single-tenant AWS Bedrock instance described above. There is no external API call to a separate AI vendor as part of generating a Decision Brief.

This matters for companies with restrictions on where financial, customer, or operational data can be sent for processing. The inference step stays within the AWS Bedrock instance assigned to that customer.

Teams evaluating SIGNLD against internal data governance policies can treat the inference step as staying inside a single, dedicated AWS Bedrock deployment rather than passing through an additional, separately governed vendor.

Is my data used to train the model

No. SIGNLD is never trained on customer data. The model used for inference is not retrained, fine-tuned, or updated using the content of any customer's connected systems or questions.

Each customer's data is used only to answer that customer's own questions through the Knowledge Graph and the single-tenant inference instance, not to improve or adjust the underlying model for anyone else.

This applies to every question type, including ones about sensitive figures such as payroll, margin, or customer risk. The answer generated for one customer's Topic has no path back into training data used for another customer.

Is every query tracked

Yes. Every query run through SIGNLD is logged and timestamped. This includes the question asked, the Topic it relates to, and the Decision Brief returned. The log creates a record of who asked what and when, across the account.

Timestamped logging supports source traceability alongside the Decision Brief itself, since each brief also links back to the specific records it drew from in connected systems.

This logging is continuous rather than optional, so an administrator reviewing account activity can see the full history of questions asked against any Topic, not just the most recent one.

How does the Knowledge Graph handle sensitive data

The Knowledge Graph resolves the same customer, job, or invoice across connected systems so a Decision Brief can draw on consistent, matched records rather than raw exports. That resolved data is stored under the same AES-256 encryption at rest as every other piece of connected data.

Access to the Knowledge Graph is scoped to the account it belongs to. Resolved entities built from one customer's connected systems are not visible to, or usable by, another customer's Topics or questions.

Who can see a Decision Brief

A Decision Brief is visible to the users on the account with access to the Topic it was generated for. The finding, evidence, confidence signal, and recommended action inside the brief travel together and are not separated or shared outside the account automatically.

Query logs record who generated or viewed a given brief, so account administrators can trace exactly which team member asked which question and when, alongside the answer that was returned.

What SIGNLD does not claim

SIGNLD does not claim zero egress, and it does not claim that data never leaves the customer's environment. Data does travel between connected source systems, the Knowledge Graph, and the single-tenant AWS Bedrock instance in order to answer a question.

What SIGNLD does maintain throughout that movement is encryption at rest and in transit, read-only source access, single-tenant AI inference, and a logged, timestamped record of every query. Read /security for the full technical documentation.

Key takeaways

• AI inference runs on a single-tenant AWS Bedrock instance.
• A Decision Brief is visible to the users on the account with access to the Topic it was generated for.
• The Knowledge Graph maintains resolved entity data drawn from connected systems to answer questions across sources.
• SIGNLD requires TLS 1.2 or higher for all data in transit.
• See the read-only connection and logging model in action on your own systems.

FAQ

Does SIGNLD store a copy of my data outside AWS Bedrock

The Knowledge Graph maintains resolved entity data drawn from connected systems to answer questions across sources. Encryption at rest with AES-256 applies to that stored data at every point.

Is TLS 1.2 the only transit protocol supported

SIGNLD requires TLS 1.2 or higher for all data in transit. Connections using older, unsupported TLS versions are not permitted.

What is a single-tenant AWS Bedrock instance

It is a private large language model deployment on AWS Bedrock dedicated to one customer's account. It is separate from a shared, multi-tenant AI service used across many customers at once.

Can a connected system be accidentally modified by SIGNLD

No. Connections are read-only, so SIGNLD has no write access to any connected system. It cannot create, edit, or delete records in a CRM, accounting platform, or any other source.

Does SIGNLD support connections to over 800 systems securely

Yes. The read-only connection model, AES-256 encryption at rest, and TLS 1.2+ in transit apply the same way across all 800-plus supported systems, regardless of which ones a given account connects.

Can I audit who asked what through SIGNLD

Yes. Every query is logged and timestamped with the Topic and Decision Brief involved, giving account administrators a full record of question activity across the account.

Where can I read the full security documentation

The complete technical security documentation, including encryption details and connection architecture, is available on /security.

Related reading in this series: SIGNLD week one: what to expect and What SIGNLD does not do.

Try SIGNLD free

See the read-only connection and logging model in action on your own systems. Try SIGNLD free or review /how-it-works for the setup steps.